2011-06-30 17:31:31 +02:00
/*
2013-02-25 15:26:00 +01:00
Copyright ( c ) 2000 , 2013 , Oracle and / or its affiliates . All rights reserved .
2000-08-30 19:40:59 +02:00
2000-07-31 21:29:14 +02:00
This program is free software ; you can redistribute it and / or modify
it under the terms of the GNU General Public License as published by
2006-12-23 20:17:15 +01:00
the Free Software Foundation ; version 2 of the License .
2000-08-30 19:40:59 +02:00
2000-07-31 21:29:14 +02:00
This program is distributed in the hope that it will be useful ,
but WITHOUT ANY WARRANTY ; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE . See the
GNU General Public License for more details .
2000-08-30 19:40:59 +02:00
2000-07-31 21:29:14 +02:00
You should have received a copy of the GNU General Public License
along with this program ; if not , write to the Free Software
2013-03-19 15:53:48 +01:00
Foundation , Inc . , 51 Franklin Street , Fifth Floor , Boston , MA 02110 - 1301 , USA
2011-06-30 17:31:31 +02:00
*/
2000-07-31 21:29:14 +02:00
/* mysqldump.c - Dump a tables contents and format to an ASCII file
* *
* * The author ' s original notes follow : -
* *
2001-10-04 19:43:06 +02:00
* * AUTHOR : Igor Romanenko ( igor @ frog . kiev . ua )
* * DATE : December 3 , 1994
* * WARRANTY : None , expressed , impressed , implied
2006-05-25 10:00:28 +02:00
* * or other
2001-10-04 19:43:06 +02:00
* * STATUS : Public domain
* * Adapted and optimized for MySQL by
* * Michael Widenius , Sinisa Milivojevic , Jani Tolonen
* * - w - - where added 9 / 10 / 98 by Jim Faucette
* * slave code by David Saez Padros < david @ ols . es >
* * master / autocommit code by Brian Aker < brian @ tangent . org >
* * SSL by
* * Andrei Errapart < andreie @ no . spam . ee >
2005-10-11 23:58:22 +02:00
* * Tõnu Samuel < tonu @ please . do . not . remove . this . spam . ee >
2001-11-05 22:48:03 +01:00
* * XML by Gary Huntress < ghuntress @ mediaone . net > 10 / 10 / 01 , cleaned up
* * and adapted to mysqldump 05 / 11 / 01 by Jani Tolonen
2002-06-11 11:37:48 +02:00
* * Added - - single - transaction option 06 / 06 / 2002 by Peter Zaitsev
2003-06-10 08:34:00 +02:00
* * 10 Jun 2003 : SET NAMES and - - no - set - names by Alexander Barkov
2001-11-05 22:48:03 +01:00
*/
2000-07-31 21:29:14 +02:00
2007-08-01 21:59:05 +02:00
# define DUMP_VERSION "10.13"
2000-07-31 21:29:14 +02:00
2001-09-14 01:54:33 +02:00
# include <my_global.h>
2000-07-31 21:29:14 +02:00
# include <my_sys.h>
2006-01-11 00:07:40 +01:00
# include <my_user.h>
2000-07-31 21:29:14 +02:00
# include <m_string.h>
# include <m_ctype.h>
2004-12-27 19:10:30 +01:00
# include <hash.h>
2006-10-19 00:43:51 +02:00
# include <stdarg.h>
2000-07-31 21:29:14 +02:00
2001-10-31 02:22:31 +01:00
# include "client_priv.h"
2000-07-31 21:29:14 +02:00
# include "mysql.h"
# include "mysql_version.h"
# include "mysqld_error.h"
2010-10-06 17:06:13 +02:00
# include <welcome_copyright_notice.h> /* ORACLE_WELCOME_COPYRIGHT_NOTICE */
2000-07-31 21:29:14 +02:00
/* Exit codes */
# define EX_USAGE 1
# define EX_MYSQLERR 2
# define EX_CONSCHECK 3
# define EX_EOM 4
2004-06-24 22:14:12 +02:00
# define EX_EOF 5 /* ferror for output file was got */
2005-06-21 14:19:56 +02:00
# define EX_ILLEGAL_TABLE 6
2000-07-31 21:29:14 +02:00
/* index into 'show fields from table' */
# define SHOW_FIELDNAME 0
# define SHOW_TYPE 1
# define SHOW_NULL 2
# define SHOW_DEFAULT 4
# define SHOW_EXTRA 5
2002-04-27 11:09:59 +02:00
/* Size of buffer for dump's select query */
# define QUERY_LENGTH 1536
2012-01-10 09:03:45 +01:00
/* Size of comment buffer. */
# define COMMENT_LENGTH 2048
2005-09-27 07:43:09 +02:00
/* ignore table flags */
# define IGNORE_NONE 0x00 /* no ignore */
# define IGNORE_DATA 0x01 /* don't dump data for this table */
# define IGNORE_INSERT_DELAYED 0x02 /* table doesn't support INSERT DELAYED */
Bug #11754178 45740: MYSQLDUMP DOESN'T DUMP GENERAL_LOG AND SLOW_QUERY
CAUSES RESTORE PROBLEM
Problem Statement:
------------------
mysqldump is not having the dump stmts for general_log and slow_log
tables. That is because of the fix for Bug#26121. Hence, after
dropping the mysql database, and applying the dump by enabling the
logging, "'general_log' table not found" errors are logged into the
server log file.
Analysis:
---------
As part of the fix for Bug#26121, we skipped the dumping of tables
for general_log and slow_log, because the data dump of those tables
are taking LOCKS, which is not allowed for log tables.
Fix:
----
We came up with an approach that instead of taking both meta data
and data dump information for those tables, take only the meta data
dump which doesn't need LOCKS.
As part of fixing the issue we came up with below algorithm.
Design before fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
Design with the fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Explicitly call the 'show create table' for general_log and
slow_log
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
While taking the meta data dump for general_log and slow_log the
"CREATE TABLE" is replaced with "CREATE TABLE IF NOT EXISTS".
This is because we skipped "DROP TABLE" for those tables,
"DROP TABLE" fails for these tables if logging is enabled.
Customer is applying the dump by enabling logging so, if the dump
has "DROP TABLE" it will fail. Hence, removed the "DROP TABLE"
stmts for those tables.
After the fix we could observe "Table 'mysql.general_log'
doesn't exist" errors initially that is because in the customer
scenario they are dropping the mysql database by enabling the
logging, Hence, those errors are expected. Once we apply the
dump which is taken before the "drop database mysql", the errors
will not be there.
2012-05-04 15:03:34 +02:00
2007-03-22 18:35:29 +01:00
static void add_load_option ( DYNAMIC_STRING * str , const char * option ,
const char * option_value ) ;
2003-01-16 04:35:59 +01:00
static ulong find_set ( TYPELIB * lib , const char * x , uint length ,
2006-05-25 10:00:28 +02:00
char * * err_pos , uint * err_len ) ;
2006-03-02 13:18:49 +01:00
static char * alloc_query_str ( ulong size ) ;
2000-07-31 21:29:14 +02:00
2007-03-22 18:35:29 +01:00
static void field_escape ( DYNAMIC_STRING * in , const char * from ) ;
2006-07-24 11:56:01 +02:00
static my_bool verbose = 0 , opt_no_create_info = 0 , opt_no_data = 0 ,
quick = 1 , extended_insert = 1 ,
2006-09-14 20:56:14 +02:00
lock_tables = 1 , ignore_errors = 0 , flush_logs = 0 , flush_privileges = 0 ,
2006-05-25 10:00:28 +02:00
opt_drop = 1 , opt_keywords = 0 , opt_lock = 1 , opt_compress = 0 ,
2003-01-16 04:35:59 +01:00
opt_delayed = 0 , create_options = 1 , opt_quoted = 0 , opt_databases = 0 ,
2005-05-18 18:25:06 +02:00
opt_alldbs = 0 , opt_create_db = 0 , opt_lock_all_tables = 0 ,
2007-10-01 17:35:51 +02:00
opt_set_charset = 0 , opt_dump_date = 1 ,
2006-05-25 10:00:28 +02:00
opt_autocommit = 0 , opt_disable_keys = 1 , opt_xml = 0 ,
opt_delete_master_logs = 0 , tty_password = 0 ,
opt_single_transaction = 0 , opt_comments = 0 , opt_compact = 0 ,
opt_hex_blob = 0 , opt_order_by_primary = 0 , opt_ignore = 0 ,
2005-07-22 04:11:45 +02:00
opt_complete_insert = 0 , opt_drop_database = 0 ,
2005-11-24 18:56:40 +01:00
opt_replace_into = 0 ,
2006-02-02 00:52:58 +01:00
opt_dump_triggers = 0 , opt_routines = 0 , opt_tz_utc = 1 ,
2009-11-04 14:31:03 +01:00
opt_slave_apply = 0 ,
opt_include_master_host_port = 0 ,
2012-01-10 09:03:45 +01:00
opt_events = 0 , opt_comments_used = 0 ,
2006-10-09 16:03:46 +02:00
opt_alltspcs = 0 , opt_notspcs = 0 ;
2007-08-01 21:59:05 +02:00
static my_bool insert_pat_inited = 0 , debug_info_flag = 0 , debug_check_flag = 0 ;
2004-05-26 18:40:27 +02:00
static ulong opt_max_allowed_packet , opt_net_buffer_length ;
2006-07-24 12:00:44 +02:00
static MYSQL mysql_connection , * mysql = 0 ;
2005-05-08 21:02:46 +02:00
static DYNAMIC_STRING insert_pat ;
static char * opt_password = 0 , * current_user = 0 ,
2000-07-31 21:29:14 +02:00
* current_host = 0 , * path = 0 , * fields_terminated = 0 ,
* lines_terminated = 0 , * enclosed = 0 , * opt_enclosed = 0 , * escaped = 0 ,
2004-11-16 23:45:24 +01:00
* where = 0 , * order_by = 0 ,
2003-03-16 08:20:45 +01:00
* opt_compatible_mode_str = 0 ,
2007-03-23 19:24:03 +01:00
* err_ptr = 0 ,
* log_error_file = NULL ;
2006-11-20 21:42:06 +01:00
static char * * defaults_argv = 0 ;
2004-03-11 15:46:27 +01:00
static char compatible_mode_normal_str [ 255 ] ;
2008-03-12 22:03:50 +01:00
/* Server supports character_set_results session variable? */
static my_bool server_supports_switching_charsets = TRUE ;
2003-01-16 04:35:59 +01:00
static ulong opt_compatible_mode = 0 ;
2004-11-10 17:56:45 +01:00
# define MYSQL_OPT_MASTER_DATA_EFFECTIVE_SQL 1
# define MYSQL_OPT_MASTER_DATA_COMMENTED_SQL 2
2009-11-04 14:31:03 +01:00
# define MYSQL_OPT_SLAVE_DATA_EFFECTIVE_SQL 1
# define MYSQL_OPT_SLAVE_DATA_COMMENTED_SQL 2
2007-08-01 21:59:05 +02:00
static uint opt_mysql_port = 0 , opt_master_data ;
2009-11-04 14:31:03 +01:00
static uint opt_slave_data ;
2007-08-01 21:59:05 +02:00
static uint my_end_arg ;
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
static char * opt_mysql_unix_port = 0 ;
2000-07-31 21:29:14 +02:00
static int first_error = 0 ;
static DYNAMIC_STRING extended_row ;
2002-09-17 22:46:53 +02:00
# include <sslopt-vars.h>
2007-03-23 19:24:03 +01:00
FILE * md_result_file = 0 ;
FILE * stderror_file = 0 ;
2002-11-14 20:16:30 +01:00
# ifdef HAVE_SMEM
static char * shared_memory_base_name = 0 ;
# endif
2003-01-16 04:35:59 +01:00
static uint opt_protocol = 0 ;
2011-01-16 04:59:05 +01:00
static char * opt_plugin_dir = 0 , * opt_default_auth = 0 ;
2007-03-22 18:35:29 +01:00
/*
Dynamic_string wrapper functions . In this file use these
wrappers , they will terminate the process if there is
an allocation failure .
*/
static void init_dynamic_string_checked ( DYNAMIC_STRING * str , const char * init_str ,
uint init_alloc , uint alloc_increment ) ;
static void dynstr_append_checked ( DYNAMIC_STRING * dest , const char * src ) ;
static void dynstr_set_checked ( DYNAMIC_STRING * str , const char * init_str ) ;
static void dynstr_append_mem_checked ( DYNAMIC_STRING * str , const char * append ,
uint length ) ;
static void dynstr_realloc_checked ( DYNAMIC_STRING * str , ulong additional_size ) ;
2005-01-06 18:13:58 +01:00
/*
2005-01-06 19:47:08 +01:00
Constant for detection of default value of default_charset .
If default_charset is equal to mysql_universal_client_charset , then
it is the default value which assigned at the very beginning of main ( ) .
2005-01-06 18:13:58 +01:00
*/
2005-01-06 19:47:08 +01:00
static const char * mysql_universal_client_charset =
2005-01-06 18:13:58 +01:00
MYSQL_UNIVERSAL_CLIENT_CHARSET ;
static char * default_charset ;
2003-03-16 08:20:45 +01:00
static CHARSET_INFO * charset_info = & my_charset_latin1 ;
2004-04-27 16:32:40 +02:00
const char * default_dbug_option = " d:t:o,/tmp/mysqldump.trace " ;
2006-05-25 10:00:28 +02:00
/* have we seen any VIEWs during table scanning? */
my_bool seen_views = 0 ;
2003-01-16 04:35:59 +01:00
const char * compatible_mode_names [ ] =
{
" MYSQL323 " , " MYSQL40 " , " POSTGRESQL " , " ORACLE " , " MSSQL " , " DB2 " ,
2004-01-15 03:26:57 +01:00
" MAXDB " , " NO_KEY_OPTIONS " , " NO_TABLE_OPTIONS " , " NO_FIELD_OPTIONS " ,
" ANSI " ,
2003-01-16 04:35:59 +01:00
NullS
} ;
2004-03-11 15:46:27 +01:00
# define MASK_ANSI_QUOTES \
( \
( 1 < < 2 ) | /* POSTGRESQL */ \
( 1 < < 3 ) | /* ORACLE */ \
( 1 < < 4 ) | /* MSSQL */ \
( 1 < < 5 ) | /* DB2 */ \
( 1 < < 6 ) | /* MAXDB */ \
( 1 < < 10 ) /* ANSI */ \
)
2003-01-16 04:35:59 +01:00
TYPELIB compatible_mode_typelib = { array_elements ( compatible_mode_names ) - 1 ,
2006-05-25 10:00:28 +02:00
" " , compatible_mode_names , NULL } ;
2003-01-16 04:35:59 +01:00
2004-12-27 19:10:30 +01:00
HASH ignore_table ;
2000-07-31 21:29:14 +02:00
2002-04-02 19:29:53 +02:00
static struct my_option my_long_options [ ] =
2000-07-31 21:29:14 +02:00
{
2002-04-02 19:29:53 +02:00
{ " all-databases " , ' A ' ,
" Dump all the databases. This will be same as --databases with all databases selected. " ,
2010-06-10 22:16:43 +02:00
& opt_alldbs , & opt_alldbs , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 ,
2002-04-02 19:29:53 +02:00
0 , 0 } ,
2006-02-02 00:52:58 +01:00
{ " all-tablespaces " , ' Y ' ,
" Dump all the tablespaces. " ,
2010-06-10 22:16:43 +02:00
& opt_alltspcs , & opt_alltspcs , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 ,
2006-02-02 00:52:58 +01:00
0 , 0 } ,
2006-10-09 16:03:46 +02:00
{ " no-tablespaces " , ' y ' ,
" Do not dump any tablespace information. " ,
2010-06-10 22:16:43 +02:00
& opt_notspcs , & opt_notspcs , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 ,
2006-10-09 16:03:46 +02:00
0 , 0 } ,
2010-02-04 13:39:42 +01:00
{ " add-drop-database " , OPT_DROP_DATABASE , " Add a DROP DATABASE before each create. " ,
2010-06-10 22:16:43 +02:00
& opt_drop_database , & opt_drop_database , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 ,
2005-05-20 15:56:02 +02:00
0 } ,
2010-02-04 13:39:42 +01:00
{ " add-drop-table " , OPT_DROP , " Add a DROP TABLE before each create. " ,
2010-06-10 22:16:43 +02:00
& opt_drop , & opt_drop , 0 , GET_BOOL , NO_ARG , 1 , 0 , 0 , 0 , 0 ,
2002-04-02 19:29:53 +02:00
0 } ,
2010-02-04 13:39:42 +01:00
{ " add-locks " , OPT_LOCKS , " Add locks around INSERT statements. " ,
2010-06-10 22:16:43 +02:00
& opt_lock , & opt_lock , 0 , GET_BOOL , NO_ARG , 1 , 0 , 0 , 0 , 0 ,
2002-04-02 19:29:53 +02:00
0 } ,
{ " allow-keywords " , OPT_KEYWORDS ,
2010-06-10 22:16:43 +02:00
" Allow creation of column names that are keywords. " , & opt_keywords ,
& opt_keywords , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2009-11-04 14:31:03 +01:00
{ " apply-slave-statements " , OPT_MYSQLDUMP_SLAVE_APPLY ,
" Adds 'STOP SLAVE' prior to 'CHANGE MASTER' and 'START SLAVE' to bottom of dump. " ,
2010-06-11 03:30:49 +02:00
& opt_slave_apply , & opt_slave_apply , 0 , GET_BOOL , NO_ARG ,
2009-11-04 14:31:03 +01:00
0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " character-sets-dir " , OPT_CHARSETS_DIR ,
2010-06-10 22:16:43 +02:00
" Directory for character set files. " , & charsets_dir ,
& charsets_dir , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2004-11-11 21:55:55 +01:00
{ " comments " , ' i ' , " Write additional information. " ,
2010-06-10 22:16:43 +02:00
& opt_comments , & opt_comments , 0 , GET_BOOL , NO_ARG ,
2004-11-11 21:55:55 +01:00
1 , 0 , 0 , 0 , 0 , 0 } ,
2003-01-16 04:35:59 +01:00
{ " compatible " , OPT_COMPATIBLE ,
2010-06-11 03:30:49 +02:00
" Change the dump to be compatible with a given mode. By default tables "
" are dumped in a format optimized for MySQL. Legal modes are: ansi, "
" mysql323, mysql40, postgresql, oracle, mssql, db2, maxdb, no_key_options, "
" no_table_options, no_field_options. One can use several modes separated "
" by commas. Note: Requires MySQL server version 4.1.0 or higher. "
" This option is ignored with earlier server versions. " ,
2010-06-10 22:16:43 +02:00
& opt_compatible_mode_str , & opt_compatible_mode_str , 0 ,
2003-01-16 04:35:59 +01:00
GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2004-02-09 12:31:03 +01:00
{ " compact " , OPT_COMPACT ,
2010-06-11 03:30:49 +02:00
" Give less verbose output (useful for debugging). Disables structure "
" comments and header/footer constructs. Enables options --skip-add- "
" drop-table --skip-add-locks --skip-comments --skip-disable-keys "
" --skip-set-charset. " ,
& opt_compact , & opt_compact , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2005-05-08 21:02:46 +02:00
{ " complete-insert " , ' c ' , " Use complete insert statements. " ,
2010-06-10 22:16:43 +02:00
& opt_complete_insert , & opt_complete_insert , 0 , GET_BOOL ,
2005-05-08 21:02:46 +02:00
NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " compress " , ' C ' , " Use compression in server/client protocol. " ,
2010-06-10 22:16:43 +02:00
& opt_compress , & opt_compress , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 ,
2002-04-02 19:29:53 +02:00
0 , 0 , 0 } ,
2010-02-17 13:15:07 +01:00
{ " create-options " , ' a ' ,
2004-04-27 14:33:40 +02:00
" Include all MySQL specific create options. " ,
2010-06-10 22:16:43 +02:00
& create_options , & create_options , 0 , GET_BOOL , NO_ARG , 1 ,
2004-04-27 14:33:40 +02:00
0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " databases " , ' B ' ,
2010-02-04 13:39:42 +01:00
" Dump several databases. Note the difference in usage; in this case no tables are given. All name arguments are regarded as database names. 'USE db_name;' will be included in the output. " ,
2010-06-10 22:16:43 +02:00
& opt_databases , & opt_databases , 0 , GET_BOOL , NO_ARG , 0 , 0 ,
2002-04-02 19:29:53 +02:00
0 , 0 , 0 , 0 } ,
2004-04-27 16:32:40 +02:00
# ifdef DBUG_OFF
2010-02-04 13:39:42 +01:00
{ " debug " , ' # ' , " This is a non-debug version. Catch this and exit. " ,
2004-04-27 16:32:40 +02:00
0 , 0 , 0 , GET_DISABLED , OPT_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
# else
2010-06-10 22:16:43 +02:00
{ " debug " , ' # ' , " Output debug log. " , & default_dbug_option ,
& default_dbug_option , 0 , GET_STR , OPT_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2004-04-27 16:32:40 +02:00
# endif
2007-10-01 14:32:07 +02:00
{ " debug-check " , OPT_DEBUG_CHECK , " Check memory and open file usage at exit. " ,
2010-06-10 22:16:43 +02:00
& debug_check_flag , & debug_check_flag , 0 ,
2007-08-01 21:59:05 +02:00
GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
{ " debug-info " , OPT_DEBUG_INFO , " Print some debug info at exit. " ,
2010-06-10 22:16:43 +02:00
& debug_info_flag , & debug_info_flag ,
2007-08-01 21:59:05 +02:00
0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " default-character-set " , OPT_DEFAULT_CHARSET ,
2010-06-10 22:16:43 +02:00
" Set the default character set. " , & default_charset ,
& default_charset , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2010-02-04 13:39:42 +01:00
{ " delayed-insert " , OPT_DELAYED , " Insert rows with INSERT DELAYED. " ,
2010-06-10 22:16:43 +02:00
& opt_delayed , & opt_delayed , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 ,
2002-04-02 19:29:53 +02:00
0 , 0 } ,
2003-04-03 11:33:13 +02:00
{ " delete-master-logs " , OPT_DELETE_MASTER_LOGS ,
2004-11-10 17:56:45 +01:00
" Delete logs on master after backup. This automatically enables --master-data. " ,
2010-06-10 22:16:43 +02:00
& opt_delete_master_logs , & opt_delete_master_logs , 0 ,
2004-11-10 17:56:45 +01:00
GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " disable-keys " , ' K ' ,
2010-06-11 03:30:49 +02:00
" '/*!40000 ALTER TABLE tb_name DISABLE KEYS */; and '/*!40000 ALTER "
" TABLE tb_name ENABLE KEYS */; will be put in the output. " , & opt_disable_keys ,
2010-06-10 22:16:43 +02:00
& opt_disable_keys , 0 , GET_BOOL , NO_ARG , 1 , 0 , 0 , 0 , 0 , 0 } ,
2009-11-04 14:31:03 +01:00
{ " dump-slave " , OPT_MYSQLDUMP_SLAVE_DATA ,
" This causes the binary log position and filename of the master to be "
" appended to the dumped data output. Setting the value to 1, will print "
" it as a CHANGE MASTER command in the dumped data output; if equal "
" to 2, that command will be prefixed with a comment symbol. "
" This option will turn --lock-all-tables on, unless "
" --single-transaction is specified too (in which case a "
" global read lock is only taken a short time at the beginning of the dump "
" - don't forget to read about --single-transaction below). In all cases "
" any action on logs will happen at the exact moment of the dump. "
" Option automatically turns --lock-tables off. " ,
2010-06-11 03:30:49 +02:00
& opt_slave_data , & opt_slave_data , 0 ,
2009-11-04 14:31:03 +01:00
GET_UINT , OPT_ARG , 0 , 0 , MYSQL_OPT_SLAVE_DATA_COMMENTED_SQL , 0 , 0 , 0 } ,
2006-03-09 21:12:43 +01:00
{ " events " , ' E ' , " Dump events. " ,
2010-06-10 22:16:43 +02:00
& opt_events , & opt_events , 0 , GET_BOOL ,
2006-03-09 21:12:43 +01:00
NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " extended-insert " , ' e ' ,
2010-01-27 16:13:39 +01:00
" Use multiple-row INSERT syntax that include several VALUES lists. " ,
2010-06-10 22:16:43 +02:00
& extended_insert , & extended_insert , 0 , GET_BOOL , NO_ARG ,
2003-01-16 04:35:59 +01:00
1 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " fields-terminated-by " , OPT_FTB ,
2010-06-11 03:30:49 +02:00
" Fields in the output file are terminated by the given string. " ,
& fields_terminated , & fields_terminated , 0 ,
2010-02-04 13:39:42 +01:00
GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " fields-enclosed-by " , OPT_ENC ,
2010-06-11 03:30:49 +02:00
" Fields in the output file are enclosed by the given character. " ,
& enclosed , & enclosed , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " fields-optionally-enclosed-by " , OPT_O_ENC ,
2010-02-24 14:52:27 +01:00
" Fields in the output file are optionally enclosed by the given character. " ,
2010-06-11 03:30:49 +02:00
& opt_enclosed , & opt_enclosed , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2010-02-24 14:52:27 +01:00
{ " fields-escaped-by " , OPT_ESC ,
2010-02-04 13:39:42 +01:00
" Fields in the output file are escaped by the given character. " ,
2010-06-11 03:30:49 +02:00
& escaped , & escaped , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2004-02-04 16:50:33 +01:00
{ " flush-logs " , ' F ' , " Flush logs file in server before starting dump. "
2004-11-10 17:56:45 +01:00
" Note that if you dump many databases at once (using the option "
" --databases= or --all-databases), the logs will be flushed for "
" each database dumped. The exception is when using --lock-all-tables "
" or --master-data: "
" in this case the logs will be flushed only once, corresponding "
" to the moment all tables are locked. So if you want your dump and "
" the log flush to happen at the same exact moment you should use "
2010-02-04 13:39:42 +01:00
" --lock-all-tables or --master-data with --flush-logs. " ,
2010-06-10 22:16:43 +02:00
& flush_logs , & flush_logs , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 ,
2002-04-02 19:29:53 +02:00
0 , 0 } ,
2006-09-14 20:56:14 +02:00
{ " flush-privileges " , OPT_ESC , " Emit a FLUSH PRIVILEGES statement "
" after dumping the mysql database. This option should be used any "
" time the dump contains the mysql database and any other database "
" that depends on the data in the mysql database for proper restore. " ,
2010-06-10 22:16:43 +02:00
& flush_privileges , & flush_privileges , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 ,
2006-09-14 20:56:14 +02:00
0 , 0 } ,
2010-02-04 13:39:42 +01:00
{ " force " , ' f ' , " Continue even if we get an SQL error. " ,
2010-06-10 22:16:43 +02:00
& ignore_errors , & ignore_errors , 0 , GET_BOOL , NO_ARG ,
2002-04-02 19:29:53 +02:00
0 , 0 , 0 , 0 , 0 , 0 } ,
{ " help " , ' ? ' , " Display this help message and exit. " , 0 , 0 , 0 , GET_NO_ARG ,
NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2004-11-11 21:55:55 +01:00
{ " hex-blob " , OPT_HEXBLOB , " Dump binary strings (BINARY, "
" VARBINARY, BLOB) in hexadecimal format. " ,
2010-06-10 22:16:43 +02:00
& opt_hex_blob , & opt_hex_blob , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
{ " host " , ' h ' , " Connect to host. " , & current_host ,
& current_host , 0 , GET_STR_ALLOC , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2004-12-27 19:10:30 +01:00
{ " ignore-table " , OPT_IGNORE_TABLE ,
" Do not dump the specified table. To specify more than one table to ignore, "
" use the directive multiple times, once for each table. Each table must "
2010-02-04 13:39:42 +01:00
" be specified with both database and table names, e.g., "
" --ignore-table=database.table. " ,
2004-12-27 19:10:30 +01:00
0 , 0 , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2009-11-04 14:31:03 +01:00
{ " include-master-host-port " , OPT_MYSQLDUMP_INCLUDE_MASTER_HOST_PORT ,
2010-06-11 03:30:49 +02:00
" Adds 'MASTER_HOST=<host>, MASTER_PORT=<port>' to 'CHANGE MASTER TO..' "
" in dump produced with --dump-slave. " , & opt_include_master_host_port ,
& opt_include_master_host_port , 0 , GET_BOOL , NO_ARG ,
2009-11-04 14:31:03 +01:00
0 , 0 , 0 , 0 , 0 , 0 } ,
2005-05-07 18:49:39 +02:00
{ " insert-ignore " , OPT_INSERT_IGNORE , " Insert rows with INSERT IGNORE. " ,
2010-06-10 22:16:43 +02:00
& opt_ignore , & opt_ignore , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 ,
2005-05-07 18:49:39 +02:00
0 , 0 } ,
2010-06-11 03:30:49 +02:00
{ " lines-terminated-by " , OPT_LTB ,
2010-02-04 13:39:42 +01:00
" Lines in the output file are terminated by the given string. " ,
2010-06-10 22:16:43 +02:00
& lines_terminated , & lines_terminated , 0 , GET_STR ,
2002-04-02 19:29:53 +02:00
REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2006-05-25 10:00:28 +02:00
{ " lock-all-tables " , ' x ' , " Locks all tables across all databases. This "
2004-11-10 17:56:45 +01:00
" is achieved by taking a global read lock for the duration of the whole "
" dump. Automatically turns --single-transaction and --lock-tables off. " ,
2010-06-10 22:16:43 +02:00
& opt_lock_all_tables , & opt_lock_all_tables , 0 , GET_BOOL , NO_ARG ,
2004-11-10 17:56:45 +01:00
0 , 0 , 0 , 0 , 0 , 0 } ,
2010-06-10 22:16:43 +02:00
{ " lock-tables " , ' l ' , " Lock all tables for read. " , & lock_tables ,
& lock_tables , 0 , GET_BOOL , NO_ARG , 1 , 0 , 0 , 0 , 0 , 0 } ,
2007-03-23 19:24:03 +01:00
{ " log-error " , OPT_ERROR_LOG_FILE , " Append warnings and errors to given file. " ,
2010-06-10 22:16:43 +02:00
& log_error_file , & log_error_file , 0 , GET_STR ,
2007-03-23 19:24:03 +01:00
REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " master-data " , OPT_MASTER_DATA ,
2004-11-10 17:56:45 +01:00
" This causes the binary log position and filename to be appended to the "
" output. If equal to 1, will print it as a CHANGE MASTER command; if equal "
" to 2, that command will be prefixed with a comment symbol. "
" This option will turn --lock-all-tables on, unless "
" --single-transaction is specified too (in which case a "
2010-02-04 13:39:42 +01:00
" global read lock is only taken a short time at the beginning of the dump; "
" don't forget to read about --single-transaction below). In all cases, "
" any action on logs will happen at the exact moment of the dump. "
2004-11-10 17:56:45 +01:00
" Option automatically turns --lock-tables off. " ,
2010-06-10 22:16:43 +02:00
& opt_master_data , & opt_master_data , 0 ,
2004-11-14 17:18:31 +01:00
GET_UINT , OPT_ARG , 0 , 0 , MYSQL_OPT_MASTER_DATA_COMMENTED_SQL , 0 , 0 , 0 } ,
2010-02-04 13:39:42 +01:00
{ " max_allowed_packet " , OPT_MAX_ALLOWED_PACKET ,
" The maximum packet length to send to or receive from server. " ,
2010-06-10 22:16:43 +02:00
& opt_max_allowed_packet , & opt_max_allowed_packet , 0 ,
2006-05-25 10:00:28 +02:00
GET_ULONG , REQUIRED_ARG , 24 * 1024 * 1024 , 4096 ,
2004-11-12 16:44:17 +01:00
( longlong ) 2L * 1024L * 1024L * 1024L , MALLOC_OVERHEAD , 1024 , 0 } ,
2010-02-04 13:39:42 +01:00
{ " net_buffer_length " , OPT_NET_BUFFER_LENGTH ,
" The buffer size for TCP/IP and socket communication. " ,
2010-06-10 22:16:43 +02:00
& opt_net_buffer_length , & opt_net_buffer_length , 0 ,
2004-11-12 16:44:17 +01:00
GET_ULONG , REQUIRED_ARG , 1024 * 1024L - 1025 , 4096 , 16 * 1024L * 1024L ,
MALLOC_OVERHEAD - 1024 , 1024 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " no-autocommit " , OPT_AUTOCOMMIT ,
" Wrap tables with autocommit/commit statements. " ,
2010-06-10 22:16:43 +02:00
& opt_autocommit , & opt_autocommit , 0 , GET_BOOL , NO_ARG ,
2002-04-02 19:29:53 +02:00
0 , 0 , 0 , 0 , 0 , 0 } ,
{ " no-create-db " , ' n ' ,
2010-02-04 13:39:42 +01:00
" Suppress the CREATE DATABASE ... IF EXISTS statement that normally is "
" output for each dumped database if --all-databases or --databases is "
" given. " ,
2010-06-10 22:16:43 +02:00
& opt_create_db , & opt_create_db , 0 ,
2010-02-04 13:39:42 +01:00
GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " no-create-info " , ' t ' , " Don't write table creation info. " ,
2010-06-10 22:16:43 +02:00
& opt_no_create_info , & opt_no_create_info , 0 , GET_BOOL ,
2006-07-24 11:56:01 +02:00
NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2010-06-10 22:16:43 +02:00
{ " no-data " , ' d ' , " No row information. " , & opt_no_data ,
& opt_no_data , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2010-06-11 03:30:49 +02:00
{ " no-set-names " , ' N ' , " Same as --skip-set-charset. " ,
2004-03-31 04:36:29 +02:00
0 , 0 , 0 , GET_NO_ARG , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " opt " , OPT_OPTIMIZE ,
2004-05-01 11:07:42 +02:00
" Same as --add-drop-table, --add-locks, --create-options, --quick, --extended-insert, --lock-tables, --set-charset, and --disable-keys. Enabled by default, disable with --skip-opt. " ,
2002-04-02 19:29:53 +02:00
0 , 0 , 0 , GET_NO_ARG , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2005-01-17 07:09:47 +01:00
{ " order-by-primary " , OPT_ORDER_BY_PRIMARY ,
" Sorts each table's rows by primary key, or first unique key, if such a key exists. Useful when dumping a MyISAM table to be loaded into an InnoDB table, but will make the dump itself take considerably longer. " ,
2010-06-10 22:16:43 +02:00
& opt_order_by_primary , & opt_order_by_primary , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " password " , ' p ' ,
" Password to use when connecting to server. If password is not given it's solicited on the tty. " ,
0 , 0 , 0 , GET_STR , OPT_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2000-07-31 21:29:14 +02:00
# ifdef __WIN__
2003-06-13 10:59:02 +02:00
{ " pipe " , ' W ' , " Use named pipes to connect to server. " , 0 , 0 , 0 , GET_NO_ARG ,
2002-04-02 19:29:53 +02:00
NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2000-07-31 21:29:14 +02:00
# endif
2010-06-10 22:16:43 +02:00
{ " port " , ' P ' , " Port number to use for connection. " , & opt_mysql_port ,
& opt_mysql_port , 0 , GET_UINT , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 ,
2002-05-11 13:36:34 +02:00
0 } ,
2010-02-04 13:39:42 +01:00
{ " protocol " , OPT_MYSQL_PROTOCOL ,
" The protocol to use for connection (tcp, socket, pipe, memory). " ,
0 , 0 , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " quick " , ' q ' , " Don't buffer query, dump directly to stdout. " ,
2010-06-10 22:16:43 +02:00
& quick , & quick , 0 , GET_BOOL , NO_ARG , 1 , 0 , 0 , 0 , 0 , 0 } ,
2003-06-13 10:59:02 +02:00
{ " quote-names " , ' Q ' , " Quote table and column names with backticks (`). " ,
2010-06-10 22:16:43 +02:00
& opt_quoted , & opt_quoted , 0 , GET_BOOL , NO_ARG , 1 , 0 , 0 , 0 ,
2002-04-02 19:29:53 +02:00
0 , 0 } ,
2005-11-29 04:44:55 +01:00
{ " replace " , OPT_MYSQL_REPLACE_INTO , " Use REPLACE INTO instead of INSERT INTO. " ,
2010-06-10 22:16:43 +02:00
& opt_replace_into , & opt_replace_into , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 ,
2005-11-24 18:56:40 +01:00
0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " result-file " , ' r ' ,
2010-07-15 13:16:06 +02:00
" Direct output to a given file. This option should be used in systems "
" (e.g., DOS, Windows) that use carriage-return linefeed pairs ( \\ r \\ n) "
" to separate text lines. This option ensures that only a single newline "
" is used. " , 0 , 0 , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2005-10-19 21:40:10 +02:00
{ " routines " , ' R ' , " Dump stored routines (functions and procedures). " ,
2010-06-11 03:30:49 +02:00
& opt_routines , & opt_routines , 0 , GET_BOOL ,
NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2004-11-12 10:17:53 +01:00
{ " set-charset " , OPT_SET_CHARSET ,
2009-11-03 10:22:22 +01:00
" Add 'SET NAMES default_character_set' to the output. " ,
2010-06-10 22:16:43 +02:00
& opt_set_charset , & opt_set_charset , 0 , GET_BOOL , NO_ARG , 1 ,
2004-11-12 10:17:53 +01:00
0 , 0 , 0 , 0 , 0 } ,
2002-11-14 20:16:30 +01:00
# ifdef HAVE_SMEM
2004-08-30 18:11:10 +02:00
{ " shared-memory-base-name " , OPT_SHARED_MEMORY_BASE_NAME ,
2010-06-10 22:16:43 +02:00
" Base name of shared memory. " , & shared_memory_base_name , & shared_memory_base_name ,
2002-11-14 20:16:30 +01:00
0 , GET_STR_ALLOC , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
# endif
2004-11-12 10:17:53 +01:00
/*
Note that the combination - - single - transaction - - master - data
will give bullet - proof binlog position only if server > = 4.1 .3 . That ' s the
old " FLUSH TABLES WITH READ LOCK does not block commit " fixed bug .
*/
2004-11-09 18:03:02 +01:00
{ " single-transaction " , OPT_TRANSACTION ,
2004-11-12 10:17:53 +01:00
" Creates a consistent snapshot by dumping all tables in a single "
" transaction. Works ONLY for tables stored in storage engines which "
" support multiversioning (currently only InnoDB does); the dump is NOT "
2007-11-20 10:25:07 +01:00
" guaranteed to be consistent for other storage engines. "
" While a --single-transaction dump is in process, to ensure a valid "
" dump file (correct table contents and binary log position), no other "
" connection should use the following statements: ALTER TABLE, DROP "
" TABLE, RENAME TABLE, TRUNCATE TABLE, as consistent snapshot is not "
" isolated from them. Option automatically turns off --lock-tables. " ,
2010-06-10 22:16:43 +02:00
& opt_single_transaction , & opt_single_transaction , 0 ,
2004-11-12 10:17:53 +01:00
GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2007-10-01 17:35:51 +02:00
{ " dump-date " , OPT_DUMP_DATE , " Put a dump date to the end of the output. " ,
2010-06-10 22:16:43 +02:00
& opt_dump_date , & opt_dump_date , 0 ,
2007-10-01 17:35:51 +02:00
GET_BOOL , NO_ARG , 1 , 0 , 0 , 0 , 0 , 0 } ,
2003-01-16 04:35:59 +01:00
{ " skip-opt " , OPT_SKIP_OPTIMIZATION ,
2004-05-01 11:07:42 +02:00
" Disable --opt. Disables --add-drop-table, --add-locks, --create-options, --quick, --extended-insert, --lock-tables, --set-charset, and --disable-keys. " ,
2003-01-16 04:35:59 +01:00
0 , 0 , 0 , GET_NO_ARG , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2010-02-04 13:39:42 +01:00
{ " socket " , ' S ' , " The socket file to use for connection. " ,
2010-06-10 22:16:43 +02:00
& opt_mysql_unix_port , & opt_mysql_unix_port , 0 ,
2010-02-04 13:39:42 +01:00
GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-09-17 22:46:53 +02:00
# include <sslopt-longopts.h>
2002-04-02 19:29:53 +02:00
{ " tab " , ' T ' ,
2010-02-04 13:39:42 +01:00
" Create tab-separated textfile for each table to given path. (Create .sql "
" and .txt files.) NOTE: This only works if mysqldump is run on the same "
" machine as the mysqld server. " ,
2010-06-10 22:16:43 +02:00
& path , & path , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " tables " , OPT_TABLES , " Overrides option --databases (-B). " ,
0 , 0 , 0 , GET_NO_ARG , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2010-02-04 13:39:42 +01:00
{ " triggers " , OPT_TRIGGERS , " Dump triggers for each dumped table. " ,
2010-06-10 22:16:43 +02:00
& opt_dump_triggers , & opt_dump_triggers , 0 , GET_BOOL ,
2010-02-04 13:39:42 +01:00
NO_ARG , 1 , 0 , 0 , 0 , 0 , 0 } ,
2005-10-13 07:44:42 +02:00
{ " tz-utc " , OPT_TZ_UTC ,
2005-10-25 19:04:31 +02:00
" SET TIME_ZONE='+00:00' at top of dump to allow dumping of TIMESTAMP data when a server has data in different time zones or data is being moved between servers with different time zones. " ,
2010-06-10 22:16:43 +02:00
& opt_tz_utc , & opt_tz_utc , 0 , GET_BOOL , NO_ARG , 1 , 0 , 0 , 0 , 0 , 0 } ,
2000-07-31 21:29:14 +02:00
# ifndef DONT_ALLOW_USER_CHANGE
2002-04-02 19:29:53 +02:00
{ " user " , ' u ' , " User for login if not current user. " ,
2010-06-10 22:16:43 +02:00
& current_user , & current_user , 0 , GET_STR , REQUIRED_ARG ,
2002-04-02 19:29:53 +02:00
0 , 0 , 0 , 0 , 0 , 0 } ,
2000-07-31 21:29:14 +02:00
# endif
2002-04-02 19:29:53 +02:00
{ " verbose " , ' v ' , " Print info about the various stages. " ,
2010-06-10 22:16:43 +02:00
& verbose , & verbose , 0 , GET_BOOL , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " version " , ' V ' , " Output version information and exit. " , 0 , 0 , 0 ,
GET_NO_ARG , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2010-02-04 13:39:42 +01:00
{ " where " , ' w ' , " Dump only selected records. Quotes are mandatory. " ,
2010-06-10 22:16:43 +02:00
& where , & where , 0 , GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2002-04-02 19:29:53 +02:00
{ " xml " , ' X ' , " Dump a database as well formed XML. " , 0 , 0 , 0 , GET_NO_ARG ,
NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2010-12-07 13:07:07 +01:00
{ " plugin_dir " , OPT_PLUGIN_DIR , " Directory for client-side plugins. " ,
2011-06-06 12:27:05 +02:00
& opt_plugin_dir , & opt_plugin_dir , 0 ,
2010-12-07 13:07:07 +01:00
GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2011-01-16 04:59:05 +01:00
{ " default_auth " , OPT_DEFAULT_AUTH ,
2010-12-07 13:07:07 +01:00
" Default authentication client-side plugin to use. " ,
2011-06-06 12:27:05 +02:00
& opt_default_auth , & opt_default_auth , 0 ,
2010-12-07 13:07:07 +01:00
GET_STR , REQUIRED_ARG , 0 , 0 , 0 , 0 , 0 , 0 } ,
2004-11-12 13:02:17 +01:00
{ 0 , 0 , 0 , 0 , 0 , 0 , GET_NO_ARG , NO_ARG , 0 , 0 , 0 , 0 , 0 , 0 }
2000-07-31 21:29:14 +02:00
} ;
static const char * load_default_groups [ ] = { " mysqldump " , " client " , 0 } ;
2007-03-22 18:35:29 +01:00
static void maybe_exit ( int error ) ;
static void die ( int error , const char * reason , . . . ) ;
static void maybe_die ( int error , const char * reason , . . . ) ;
2002-01-02 20:29:41 +01:00
static void write_header ( FILE * sql_file , char * db_name ) ;
2000-07-31 21:29:14 +02:00
static void print_value ( FILE * file , MYSQL_RES * result , MYSQL_ROW row ,
2006-05-25 10:00:28 +02:00
const char * prefix , const char * name ,
int string_value ) ;
2000-07-31 21:29:14 +02:00
static int dump_selected_tables ( char * db , char * * table_names , int tables ) ;
static int dump_all_tables_in_db ( char * db ) ;
2006-08-28 23:48:06 +02:00
static int init_dumping_views ( char * ) ;
static int init_dumping_tables ( char * ) ;
static int init_dumping ( char * , int init_func ( char * ) ) ;
2000-07-31 21:29:14 +02:00
static int dump_databases ( char * * ) ;
static int dump_all_databases ( ) ;
2003-09-03 17:48:10 +02:00
static char * quote_name ( const char * name , char * buff , my_bool force ) ;
2005-09-27 07:43:09 +02:00
char check_if_ignore_table ( const char * table_name , char * table_type ) ;
2004-11-16 23:45:24 +01:00
static char * primary_key_fields ( const char * table_name ) ;
2005-04-04 23:32:48 +02:00
static my_bool get_view_structure ( char * table , char * db ) ;
2004-11-12 04:01:46 +01:00
static my_bool dump_all_views_in_db ( char * database ) ;
2006-10-09 16:03:46 +02:00
static int dump_all_tablespaces ( ) ;
static int dump_tablespaces_for_tables ( char * db , char * * table_names , int tables ) ;
static int dump_tablespaces_for_databases ( char * * databases ) ;
static int dump_tablespaces ( char * ts_where ) ;
2012-01-10 09:03:45 +01:00
static void print_comment ( FILE * sql_file , my_bool is_error , const char * format ,
. . . ) ;
2003-12-14 05:39:52 +01:00
2000-07-31 21:29:14 +02:00
2006-07-24 11:51:52 +02:00
/*
Print the supplied message if in verbose mode
SYNOPSIS
verbose_msg ( )
fmt format specifier
. . . variable number of parameters
*/
static void verbose_msg ( const char * fmt , . . . )
{
va_list args ;
DBUG_ENTER ( " verbose_msg " ) ;
if ( ! verbose )
DBUG_VOID_RETURN ;
va_start ( args , fmt ) ;
vfprintf ( stderr , fmt , args ) ;
va_end ( args ) ;
2011-12-24 10:38:59 +01:00
fflush ( stderr ) ;
2006-07-24 11:51:52 +02:00
DBUG_VOID_RETURN ;
}
2004-06-24 22:14:12 +02:00
/*
exit with message if ferror ( file )
2006-05-25 10:00:28 +02:00
2004-06-24 22:14:12 +02:00
SYNOPSIS
check_io ( )
2006-05-25 10:00:28 +02:00
file - checked file
2004-06-24 22:14:12 +02:00
*/
void check_io ( FILE * file )
{
if ( ferror ( file ) )
2007-03-22 18:35:29 +01:00
die ( EX_EOF , " Got errno %d on write " , errno ) ;
2004-06-24 22:14:12 +02:00
}
2000-07-31 21:29:14 +02:00
static void print_version ( void )
{
printf ( " %s Ver %s Distrib %s, for %s (%s) \n " , my_progname , DUMP_VERSION ,
2004-05-25 21:00:14 +02:00
MYSQL_SERVER_VERSION , SYSTEM_TYPE , MACHINE_TYPE ) ;
2000-07-31 21:29:14 +02:00
} /* print_version */
2002-09-24 16:11:59 +02:00
static void short_usage_sub ( void )
{
printf ( " Usage: %s [OPTIONS] database [tables] \n " , my_progname ) ;
printf ( " OR %s [OPTIONS] --databases [OPTIONS] DB1 [DB2 DB3...] \n " ,
2006-05-25 10:00:28 +02:00
my_progname ) ;
2002-09-24 16:11:59 +02:00
printf ( " OR %s [OPTIONS] --all-databases [OPTIONS] \n " , my_progname ) ;
}
2004-05-25 21:00:14 +02:00
2000-07-31 21:29:14 +02:00
static void usage ( void )
{
print_version ( ) ;
2012-08-07 15:28:19 +02:00
puts ( ORACLE_WELCOME_COPYRIGHT_NOTICE ( " 2000 " ) ) ;
2010-02-04 13:39:42 +01:00
puts ( " Dumping structure and contents of MySQL databases and tables. " ) ;
2002-09-24 16:11:59 +02:00
short_usage_sub ( ) ;
2000-07-31 21:29:14 +02:00
print_defaults ( " my " , load_default_groups ) ;
2002-04-02 19:29:53 +02:00
my_print_help ( my_long_options ) ;
my_print_variables ( my_long_options ) ;
2000-07-31 21:29:14 +02:00
} /* usage */
2002-09-24 16:11:59 +02:00
static void short_usage ( void )
{
short_usage_sub ( ) ;
printf ( " For more options, use %s --help \n " , my_progname ) ;
}
2002-01-02 20:29:41 +01:00
static void write_header ( FILE * sql_file , char * db_name )
2000-07-31 21:29:14 +02:00
{
2001-11-05 22:48:03 +01:00
if ( opt_xml )
2002-11-18 17:50:42 +01:00
{
2003-10-30 11:58:30 +01:00
fputs ( " <?xml version= \" 1.0 \" ?> \n " , sql_file ) ;
2006-11-27 00:47:38 +01:00
/*
Schema reference . Allows use of xsi : nil for NULL values and
xsi : type to define an element ' s data type .
*/
2005-06-22 20:22:54 +02:00
fputs ( " <mysqldump " , sql_file ) ;
fputs ( " xmlns:xsi= \" http://www.w3.org/2001/XMLSchema-instance \" " ,
sql_file ) ;
fputs ( " > \n " , sql_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( sql_file ) ;
2002-11-18 17:50:42 +01:00
}
2004-02-09 12:31:03 +01:00
else if ( ! opt_compact )
2001-11-05 22:48:03 +01:00
{
2012-01-10 09:03:45 +01:00
print_comment ( sql_file , 0 ,
" -- MySQL dump %s Distrib %s, for %s (%s) \n -- \n " ,
DUMP_VERSION , MYSQL_SERVER_VERSION , SYSTEM_TYPE ,
MACHINE_TYPE ) ;
print_comment ( sql_file , 0 , " -- Host: %s Database: %s \n " ,
current_host ? current_host : " localhost " ,
db_name ? db_name : " " ) ;
print_comment ( sql_file , 0 ,
" -- ------------------------------------------------------ \n "
) ;
print_comment ( sql_file , 0 , " -- Server version \t %s \n " ,
mysql_get_server_info ( & mysql_connection ) ) ;
2004-03-31 04:36:29 +02:00
if ( opt_set_charset )
2004-06-07 15:12:23 +02:00
fprintf ( sql_file ,
" \n /*!40101 SET @OLD_CHARACTER_SET_CLIENT=@@CHARACTER_SET_CLIENT */; "
" \n /*!40101 SET @OLD_CHARACTER_SET_RESULTS=@@CHARACTER_SET_RESULTS */; "
" \n /*!40101 SET @OLD_COLLATION_CONNECTION=@@COLLATION_CONNECTION */; "
" \n /*!40101 SET NAMES %s */; \n " , default_charset ) ;
2005-10-13 07:44:42 +02:00
if ( opt_tz_utc )
{
fprintf ( sql_file , " /*!40103 SET @OLD_TIME_ZONE=@@TIME_ZONE */; \n " ) ;
fprintf ( sql_file , " /*!40103 SET TIME_ZONE='+00:00' */; \n " ) ;
}
2004-02-13 22:21:46 +01:00
if ( ! path )
{
fprintf ( md_result_file , " \
2003-07-23 00:58:30 +02:00
/*!40014 SET @OLD_UNIQUE_CHECKS=@@UNIQUE_CHECKS, UNIQUE_CHECKS=0 */ ; \ n \
/*!40014 SET @OLD_FOREIGN_KEY_CHECKS=@@FOREIGN_KEY_CHECKS, FOREIGN_KEY_CHECKS=0 */ ; \ n \
" );
2004-02-13 22:21:46 +01:00
}
fprintf ( sql_file ,
2006-05-25 10:00:28 +02:00
" /*!40101 SET @OLD_SQL_MODE=@@SQL_MODE, SQL_MODE='%s%s%s' */; \n "
" /*!40111 SET @OLD_SQL_NOTES=@@SQL_NOTES, SQL_NOTES=0 */; \n " ,
path ? " " : " NO_AUTO_VALUE_ON_ZERO " , compatible_mode_normal_str [ 0 ] = = 0 ? " " : " , " ,
compatible_mode_normal_str ) ;
2004-06-24 22:14:12 +02:00
check_io ( sql_file ) ;
2001-11-05 22:48:03 +01:00
}
2002-01-02 20:29:41 +01:00
} /* write_header */
2000-07-31 21:29:14 +02:00
2003-01-16 04:35:59 +01:00
2002-11-18 17:50:42 +01:00
static void write_footer ( FILE * sql_file )
{
if ( opt_xml )
2004-06-24 22:14:12 +02:00
{
2004-02-09 12:31:03 +01:00
fputs ( " </mysqldump> \n " , sql_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( sql_file ) ;
}
2004-02-09 12:31:03 +01:00
else if ( ! opt_compact )
2003-07-23 00:58:30 +02:00
{
2005-10-13 07:44:42 +02:00
if ( opt_tz_utc )
fprintf ( sql_file , " /*!40103 SET TIME_ZONE=@OLD_TIME_ZONE */; \n " ) ;
2004-02-13 22:21:46 +01:00
fprintf ( sql_file , " \n /*!40101 SET SQL_MODE=@OLD_SQL_MODE */; \n " ) ;
if ( ! path )
{
fprintf ( md_result_file , " \
2003-07-23 00:58:30 +02:00
/*!40014 SET FOREIGN_KEY_CHECKS=@OLD_FOREIGN_KEY_CHECKS */ ; \ n \
2004-02-09 12:31:03 +01:00
/*!40014 SET UNIQUE_CHECKS=@OLD_UNIQUE_CHECKS */ ; \ n " );
2004-02-13 22:21:46 +01:00
}
2004-03-31 04:36:29 +02:00
if ( opt_set_charset )
2004-02-13 22:21:46 +01:00
fprintf ( sql_file ,
2004-06-07 15:12:23 +02:00
" /*!40101 SET CHARACTER_SET_CLIENT=@OLD_CHARACTER_SET_CLIENT */; \n "
" /*!40101 SET CHARACTER_SET_RESULTS=@OLD_CHARACTER_SET_RESULTS */; \n "
" /*!40101 SET COLLATION_CONNECTION=@OLD_COLLATION_CONNECTION */; \n " ) ;
2005-02-21 18:40:28 +01:00
fprintf ( sql_file ,
2006-05-25 10:00:28 +02:00
" /*!40111 SET SQL_NOTES=@OLD_SQL_NOTES */; \n " ) ;
2004-02-09 12:31:03 +01:00
fputs ( " \n " , sql_file ) ;
2012-01-10 09:03:45 +01:00
if ( opt_dump_date )
2006-07-24 10:56:33 +02:00
{
2012-01-10 09:03:45 +01:00
char time_str [ 20 ] ;
get_date ( time_str , GETDATE_DATE_TIME , 0 ) ;
print_comment ( sql_file , 0 , " -- Dump completed on %s \n " , time_str ) ;
2006-07-24 10:56:33 +02:00
}
2012-01-10 09:03:45 +01:00
else
print_comment ( sql_file , 0 , " -- Dump completed \n " ) ;
2004-06-24 22:14:12 +02:00
check_io ( sql_file ) ;
2003-07-23 00:58:30 +02:00
}
2002-11-18 17:50:42 +01:00
} /* write_footer */
2000-07-31 21:29:14 +02:00
2005-06-27 19:31:00 +02:00
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
uchar * get_table_key ( const char * entry , size_t * length ,
my_bool not_used __attribute__ ( ( unused ) ) )
2004-12-27 19:10:30 +01:00
{
2005-06-21 14:19:56 +02:00
* length = strlen ( entry ) ;
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
return ( uchar * ) entry ;
2004-12-27 19:10:30 +01:00
}
2002-04-02 19:29:53 +02:00
static my_bool
get_one_option ( int optid , const struct my_option * opt __attribute__ ( ( unused ) ) ,
2006-05-25 10:00:28 +02:00
char * argument )
2000-07-31 21:29:14 +02:00
{
2003-04-03 11:33:13 +02:00
switch ( optid ) {
2002-04-02 19:29:53 +02:00
case ' p ' :
2009-05-07 19:51:55 +02:00
if ( argument = = disabled_my_option )
2009-06-03 13:18:12 +02:00
argument = ( char * ) " " ; /* Don't require password */
2002-04-02 19:29:53 +02:00
if ( argument )
{
char * start = argument ;
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( opt_password ) ;
2002-04-02 19:29:53 +02:00
opt_password = my_strdup ( argument , MYF ( MY_FAE ) ) ;
2006-05-25 10:00:28 +02:00
while ( * argument ) * argument + + = ' x ' ; /* Destroy argument */
2002-04-02 19:29:53 +02:00
if ( * start )
2006-05-25 10:00:28 +02:00
start [ 1 ] = 0 ; /* Cut length of argument */
2004-10-07 21:08:17 +02:00
tty_password = 0 ;
2002-04-02 19:29:53 +02:00
}
else
tty_password = 1 ;
break ;
case ' r ' :
2006-08-02 03:18:47 +02:00
if ( ! ( md_result_file = my_fopen ( argument , O_WRONLY | FILE_BINARY ,
2006-05-25 10:00:28 +02:00
MYF ( MY_WME ) ) ) )
2002-04-02 19:29:53 +02:00
exit ( 1 ) ;
break ;
case ' W ' :
2000-07-31 21:29:14 +02:00
# ifdef __WIN__
2006-08-02 03:18:47 +02:00
opt_protocol = MYSQL_PROTOCOL_PIPE ;
2000-07-31 21:29:14 +02:00
# endif
2002-04-02 19:29:53 +02:00
break ;
2004-03-31 04:36:29 +02:00
case ' N ' :
opt_set_charset = 0 ;
break ;
2002-04-02 19:29:53 +02:00
case ' T ' :
opt_disable_keys = 0 ;
2007-06-28 11:23:59 +02:00
if ( strlen ( argument ) > = FN_REFLEN )
{
/*
This check is made because the some the file functions below
have FN_REFLEN sized stack allocated buffers and will cause
a crash even if the input destination buffer is large enough
to hold the output .
*/
die ( EX_USAGE , " Input filename too long: %s " , argument ) ;
}
2002-04-02 19:29:53 +02:00
break ;
case ' # ' :
2004-04-27 16:32:40 +02:00
DBUG_PUSH ( argument ? argument : default_dbug_option ) ;
2007-08-02 06:49:29 +02:00
debug_check_flag = 1 ;
2002-04-02 19:29:53 +02:00
break ;
2002-09-17 22:46:53 +02:00
# include <sslopt-case.h>
2002-04-02 19:29:53 +02:00
case ' V ' : print_version ( ) ; exit ( 0 ) ;
case ' X ' :
2006-08-02 03:18:47 +02:00
opt_xml = 1 ;
2006-05-25 10:00:28 +02:00
extended_insert = opt_drop = opt_lock =
2003-10-14 12:19:54 +02:00
opt_disable_keys = opt_autocommit = opt_create_db = 0 ;
2002-04-02 19:29:53 +02:00
break ;
2012-01-10 09:03:45 +01:00
case ' i ' :
opt_comments_used = 1 ;
break ;
2002-04-02 19:29:53 +02:00
case ' I ' :
case ' ? ' :
usage ( ) ;
exit ( 0 ) ;
2004-11-14 17:18:31 +01:00
case ( int ) OPT_MASTER_DATA :
if ( ! argument ) /* work like in old versions */
opt_master_data = MYSQL_OPT_MASTER_DATA_EFFECTIVE_SQL ;
break ;
2009-11-04 14:31:03 +01:00
case ( int ) OPT_MYSQLDUMP_SLAVE_DATA :
if ( ! argument ) /* work like in old versions */
opt_slave_data = MYSQL_OPT_SLAVE_DATA_EFFECTIVE_SQL ;
break ;
2002-04-02 19:29:53 +02:00
case ( int ) OPT_OPTIMIZE :
2003-01-16 04:35:59 +01:00
extended_insert = opt_drop = opt_lock = quick = create_options =
2004-03-31 04:36:29 +02:00
opt_disable_keys = lock_tables = opt_set_charset = 1 ;
2002-04-02 19:29:53 +02:00
break ;
2003-01-16 04:35:59 +01:00
case ( int ) OPT_SKIP_OPTIMIZATION :
extended_insert = opt_drop = opt_lock = quick = create_options =
2004-04-27 14:33:40 +02:00
opt_disable_keys = lock_tables = opt_set_charset = 0 ;
2003-01-16 04:35:59 +01:00
break ;
2004-02-09 12:31:03 +01:00
case ( int ) OPT_COMPACT :
2012-01-10 09:03:45 +01:00
if ( opt_compact )
{
opt_comments = opt_drop = opt_disable_keys = opt_lock = 0 ;
opt_set_charset = 0 ;
}
break ;
2002-04-02 19:29:53 +02:00
case ( int ) OPT_TABLES :
opt_databases = 0 ;
break ;
2004-12-27 19:10:30 +01:00
case ( int ) OPT_IGNORE_TABLE :
{
2004-12-28 17:33:49 +01:00
if ( ! strchr ( argument , ' . ' ) )
2004-12-27 19:10:30 +01:00
{
2004-12-28 17:33:49 +01:00
fprintf ( stderr , " Illegal use of option --ignore-table=<database>.<table> \n " ) ;
2004-12-27 19:10:30 +01:00
exit ( 1 ) ;
}
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
if ( my_hash_insert ( & ignore_table , ( uchar * ) my_strdup ( argument , MYF ( 0 ) ) ) )
2004-12-28 17:33:49 +01:00
exit ( EX_EOM ) ;
2004-12-27 19:10:30 +01:00
break ;
}
2003-01-16 04:35:59 +01:00
case ( int ) OPT_COMPATIBLE :
2004-12-28 17:33:49 +01:00
{
2003-01-16 04:35:59 +01:00
char buff [ 255 ] ;
2004-03-11 15:46:27 +01:00
char * end = compatible_mode_normal_str ;
int i ;
ulong mode ;
2006-12-14 23:51:37 +01:00
uint err_len ;
2003-01-16 04:35:59 +01:00
opt_quoted = 1 ;
2004-03-31 04:36:29 +02:00
opt_set_charset = 0 ;
2003-01-16 04:35:59 +01:00
opt_compatible_mode_str = argument ;
opt_compatible_mode = find_set ( & compatible_mode_typelib ,
2009-02-10 23:47:54 +01:00
argument , ( uint ) strlen ( argument ) ,
2006-05-25 10:00:28 +02:00
& err_ptr , & err_len ) ;
2003-01-16 04:35:59 +01:00
if ( err_len )
{
2009-12-17 18:58:38 +01:00
strmake ( buff , err_ptr , min ( sizeof ( buff ) - 1 , err_len ) ) ;
2006-05-25 10:00:28 +02:00
fprintf ( stderr , " Invalid mode to --compatible: %s \n " , buff ) ;
exit ( 1 ) ;
2003-01-16 04:35:59 +01:00
}
2004-03-11 15:46:27 +01:00
# if !defined(DBUG_OFF)
{
2006-05-25 10:00:28 +02:00
uint size_for_sql_mode = 0 ;
const char * * ptr ;
for ( ptr = compatible_mode_names ; * ptr ; ptr + + )
size_for_sql_mode + = strlen ( * ptr ) ;
size_for_sql_mode + = sizeof ( compatible_mode_names ) - 1 ;
DBUG_ASSERT ( sizeof ( compatible_mode_normal_str ) > = size_for_sql_mode ) ;
2004-03-11 15:46:27 +01:00
}
# endif
mode = opt_compatible_mode ;
for ( i = 0 , mode = opt_compatible_mode ; mode ; mode > > = 1 , i + + )
{
2006-05-25 10:00:28 +02:00
if ( mode & 1 )
{
end = strmov ( end , compatible_mode_names [ i ] ) ;
end = strmov ( end , " , " ) ;
}
2004-03-11 15:46:27 +01:00
}
if ( end ! = compatible_mode_normal_str )
2006-05-25 10:00:28 +02:00
end [ - 1 ] = 0 ;
/*
2004-12-22 13:02:27 +01:00
Set charset to the default compiled value if it hasn ' t
been reset yet by - - default - character - set = xxx .
*/
2005-01-06 18:13:58 +01:00
if ( default_charset = = mysql_universal_client_charset )
2004-12-22 13:02:27 +01:00
default_charset = ( char * ) MYSQL_DEFAULT_CHARSET_NAME ;
2003-01-16 04:35:59 +01:00
break ;
}
case ( int ) OPT_MYSQL_PROTOCOL :
2007-03-19 10:19:51 +01:00
opt_protocol = find_type_or_exit ( argument , & sql_protocol_typelib ,
opt - > name ) ;
break ;
2000-07-31 21:29:14 +02:00
}
2002-04-02 19:29:53 +02:00
return 0 ;
}
static int get_options ( int * argc , char * * * argv )
{
int ho_error ;
2004-05-26 18:40:27 +02:00
MYSQL_PARAMETERS * mysql_params = mysql_get_parameters ( ) ;
opt_max_allowed_packet = * mysql_params - > p_max_allowed_packet ;
opt_net_buffer_length = * mysql_params - > p_net_buffer_length ;
2002-04-02 19:29:53 +02:00
md_result_file = stdout ;
2009-10-09 16:44:22 +02:00
if ( load_defaults ( " my " , load_default_groups , argc , argv ) )
return 1 ;
2006-11-20 21:42:06 +01:00
defaults_argv = * argv ;
2002-04-02 19:29:53 +02:00
2009-10-14 18:37:38 +02:00
if ( my_hash_init ( & ignore_table , charset_info , 16 , 0 , 0 ,
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
( my_hash_get_key ) get_table_key , my_free , 0 ) )
2006-11-20 21:42:06 +01:00
return ( EX_EOM ) ;
2007-04-05 21:53:02 +02:00
/* Don't copy internal log tables */
2006-11-20 21:42:06 +01:00
if ( my_hash_insert ( & ignore_table ,
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
( uchar * ) my_strdup ( " mysql.apply_status " , MYF ( MY_WME ) ) ) | |
2006-11-20 21:42:06 +01:00
my_hash_insert ( & ignore_table ,
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
( uchar * ) my_strdup ( " mysql.schema " , MYF ( MY_WME ) ) ) | |
2007-04-05 21:53:02 +02:00
my_hash_insert ( & ignore_table ,
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
( uchar * ) my_strdup ( " mysql.general_log " , MYF ( MY_WME ) ) ) | |
2007-04-05 21:53:02 +02:00
my_hash_insert ( & ignore_table ,
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
( uchar * ) my_strdup ( " mysql.slow_log " , MYF ( MY_WME ) ) ) )
2006-11-20 21:42:06 +01:00
return ( EX_EOM ) ;
2002-04-02 19:29:53 +02:00
2006-11-20 21:42:06 +01:00
if ( ( ho_error = handle_options ( argc , argv , my_long_options , get_one_option ) ) )
return ( ho_error ) ;
2002-04-02 19:29:53 +02:00
2004-05-26 18:40:27 +02:00
* mysql_params - > p_max_allowed_packet = opt_max_allowed_packet ;
* mysql_params - > p_net_buffer_length = opt_net_buffer_length ;
2007-08-01 21:59:05 +02:00
if ( debug_info_flag )
my_end_arg = MY_CHECK_ERROR | MY_GIVE_INFO ;
if ( debug_check_flag )
my_end_arg = MY_CHECK_ERROR ;
2004-05-26 18:40:27 +02:00
2000-07-31 21:29:14 +02:00
if ( opt_delayed )
2006-05-25 10:00:28 +02:00
opt_lock = 0 ; /* Can't have lock with delayed */
2000-07-31 21:29:14 +02:00
if ( ! path & & ( enclosed | | opt_enclosed | | escaped | | lines_terminated | |
2006-05-25 10:00:28 +02:00
fields_terminated ) )
2000-07-31 21:29:14 +02:00
{
2002-04-02 19:29:53 +02:00
fprintf ( stderr ,
2006-05-25 10:00:28 +02:00
" %s: You must use option --tab with --fields-... \n " , my_progname ) ;
2006-11-20 21:42:06 +01:00
return ( EX_USAGE ) ;
2000-07-31 21:29:14 +02:00
}
2004-11-10 17:56:45 +01:00
2009-11-04 14:31:03 +01:00
/* We don't delete master logs if slave data option */
if ( opt_slave_data )
{
opt_lock_all_tables = ! opt_single_transaction ;
opt_master_data = 0 ;
opt_delete_master_logs = 0 ;
}
2004-11-10 17:56:45 +01:00
/* Ensure consistency of the set of binlog & locking options */
if ( opt_delete_master_logs & & ! opt_master_data )
opt_master_data = MYSQL_OPT_MASTER_DATA_COMMENTED_SQL ;
if ( opt_single_transaction & & opt_lock_all_tables )
{
fprintf ( stderr , " %s: You can't use --single-transaction and "
" --lock-all-tables at the same time. \n " , my_progname ) ;
2006-11-20 21:42:06 +01:00
return ( EX_USAGE ) ;
2006-05-25 10:00:28 +02:00
}
2004-11-10 17:56:45 +01:00
if ( opt_master_data )
2009-11-04 14:31:03 +01:00
{
2004-11-10 17:56:45 +01:00
opt_lock_all_tables = ! opt_single_transaction ;
2009-11-04 14:31:03 +01:00
opt_slave_data = 0 ;
}
2004-11-10 17:56:45 +01:00
if ( opt_single_transaction | | opt_lock_all_tables )
2003-01-16 04:35:59 +01:00
lock_tables = 0 ;
2000-07-31 21:29:14 +02:00
if ( enclosed & & opt_enclosed )
{
fprintf ( stderr , " %s: You can't use ..enclosed.. and ..optionally-enclosed.. at the same time. \n " , my_progname ) ;
2006-11-20 21:42:06 +01:00
return ( EX_USAGE ) ;
2000-07-31 21:29:14 +02:00
}
if ( ( opt_databases | | opt_alldbs ) & & path )
{
fprintf ( stderr ,
2006-05-25 10:00:28 +02:00
" %s: --databases or --all-databases can't be used with --tab. \n " ,
my_progname ) ;
2006-11-20 21:42:06 +01:00
return ( EX_USAGE ) ;
2000-07-31 21:29:14 +02:00
}
2004-02-11 12:15:39 +01:00
if ( strcmp ( default_charset , charset_info - > csname ) & &
2006-05-25 10:00:28 +02:00
! ( charset_info = get_charset_by_csname ( default_charset ,
MY_CS_PRIMARY , MYF ( MY_WME ) ) ) )
2003-03-16 08:20:45 +01:00
exit ( 1 ) ;
2000-07-31 21:29:14 +02:00
if ( ( * argc < 1 & & ! opt_alldbs ) | | ( * argc > 0 & & opt_alldbs ) )
{
2002-09-24 16:11:59 +02:00
short_usage ( ) ;
2006-11-20 21:42:06 +01:00
return EX_USAGE ;
2000-07-31 21:29:14 +02:00
}
if ( tty_password )
2000-11-13 22:55:10 +01:00
opt_password = get_tty_password ( NullS ) ;
2000-07-31 21:29:14 +02:00
return ( 0 ) ;
} /* get_options */
/*
2005-04-04 23:32:48 +02:00
* * DB_error - - prints mysql error message and exits the program .
2000-07-31 21:29:14 +02:00
*/
2006-12-14 23:51:37 +01:00
static void DB_error ( MYSQL * mysql_arg , const char * when )
2000-07-31 21:29:14 +02:00
{
2005-04-04 23:32:48 +02:00
DBUG_ENTER ( " DB_error " ) ;
2007-03-22 18:35:29 +01:00
maybe_die ( EX_MYSQLERR , " Got error: %d: %s %s " ,
2006-12-14 23:51:37 +01:00
mysql_errno ( mysql_arg ) , mysql_error ( mysql_arg ) , when ) ;
2000-07-31 21:29:14 +02:00
DBUG_VOID_RETURN ;
2007-03-22 18:35:29 +01:00
}
/*
Prints out an error message and kills the process .
SYNOPSIS
die ( )
error_num - process return value
fmt_reason - a format string for use by my_vsnprintf .
. . . - variable arguments for above fmt_reason string
DESCRIPTION
This call prints out the formatted error message to stderr and then
terminates the process .
*/
static void die ( int error_num , const char * fmt_reason , . . . )
{
char buffer [ 1000 ] ;
va_list args ;
va_start ( args , fmt_reason ) ;
my_vsnprintf ( buffer , sizeof ( buffer ) , fmt_reason , args ) ;
va_end ( args ) ;
fprintf ( stderr , " %s: %s \n " , my_progname , buffer ) ;
fflush ( stderr ) ;
ignore_errors = 0 ; /* force the exit */
maybe_exit ( error_num ) ;
}
/*
Prints out an error message and maybe kills the process .
SYNOPSIS
maybe_die ( )
error_num - process return value
fmt_reason - a format string for use by my_vsnprintf .
. . . - variable arguments for above fmt_reason string
DESCRIPTION
This call prints out the formatted error message to stderr and then
terminates the process , unless the - - force command line option is used .
This call should be used for non - fatal errors ( such as database
errors ) that the code may still be able to continue to the next unit
of work .
*/
static void maybe_die ( int error_num , const char * fmt_reason , . . . )
{
char buffer [ 1000 ] ;
va_list args ;
va_start ( args , fmt_reason ) ;
my_vsnprintf ( buffer , sizeof ( buffer ) , fmt_reason , args ) ;
va_end ( args ) ;
fprintf ( stderr , " %s: %s \n " , my_progname , buffer ) ;
fflush ( stderr ) ;
maybe_exit ( error_num ) ;
}
2000-07-31 21:29:14 +02:00
2004-11-10 17:56:45 +01:00
/*
Sends a query to server , optionally reads result , prints error message if
some .
SYNOPSIS
mysql_query_with_error_report ( )
mysql_con connection to use
2005-07-03 13:17:52 +02:00
res if non zero , result will be put there with
2006-05-25 10:00:28 +02:00
mysql_store_result ( )
2004-11-10 17:56:45 +01:00
query query to send to server
RETURN VALUES
0 query sending and ( if res ! = 0 ) result reading went ok
1 error
*/
2006-05-25 10:00:28 +02:00
2004-11-10 17:56:45 +01:00
static int mysql_query_with_error_report ( MYSQL * mysql_con , MYSQL_RES * * res ,
const char * query )
{
if ( mysql_query ( mysql_con , query ) | |
( res & & ! ( ( * res ) = mysql_store_result ( mysql_con ) ) ) )
{
2007-03-22 18:35:29 +01:00
maybe_die ( EX_MYSQLERR , " Couldn't execute '%s': %s (%d) " ,
query , mysql_error ( mysql_con ) , mysql_errno ( mysql_con ) ) ;
2004-11-10 17:56:45 +01:00
return 1 ;
}
return 0 ;
}
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
static int fetch_db_collation ( const char * db_name ,
char * db_cl_name ,
int db_cl_size )
{
2008-02-18 23:29:39 +01:00
my_bool err_status = FALSE ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
char query [ QUERY_LENGTH ] ;
MYSQL_RES * db_cl_res ;
MYSQL_ROW db_cl_row ;
2007-11-02 09:24:45 +01:00
char quoted_database_buf [ NAME_LEN * 2 + 3 ] ;
char * qdatabase = quote_name ( db_name , quoted_database_buf , 1 ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-11-02 09:24:45 +01:00
my_snprintf ( query , sizeof ( query ) , " use %s " , qdatabase ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
if ( mysql_query_with_error_report ( mysql , NULL , query ) )
return 1 ;
if ( mysql_query_with_error_report ( mysql , & db_cl_res ,
" select @@collation_database " ) )
return 1 ;
do
{
if ( mysql_num_rows ( db_cl_res ) ! = 1 )
{
err_status = TRUE ;
break ;
}
if ( ! ( db_cl_row = mysql_fetch_row ( db_cl_res ) ) )
{
err_status = TRUE ;
break ;
}
strncpy ( db_cl_name , db_cl_row [ 0 ] , db_cl_size ) ;
db_cl_name [ db_cl_size - 1 ] = 0 ; /* just in case. */
} while ( FALSE ) ;
mysql_free_result ( db_cl_res ) ;
return err_status ? 1 : 0 ;
}
static char * my_case_str ( const char * str ,
uint str_len ,
const char * token ,
uint token_len )
{
my_match_t match ;
uint status = my_charset_latin1 . coll - > instr ( & my_charset_latin1 ,
str , str_len ,
token , token_len ,
& match , 1 ) ;
return status ? ( char * ) str + match . end : NULL ;
}
static int switch_db_collation ( FILE * sql_file ,
const char * db_name ,
const char * delimiter ,
const char * current_db_cl_name ,
2007-06-29 14:52:05 +02:00
const char * required_db_cl_name ,
int * db_cl_altered )
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
{
if ( strcmp ( current_db_cl_name , required_db_cl_name ) ! = 0 )
{
2011-02-21 08:07:24 +01:00
char quoted_db_buf [ NAME_LEN * 2 + 3 ] ;
char * quoted_db_name = quote_name ( db_name , quoted_db_buf , FALSE ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
CHARSET_INFO * db_cl = get_charset_by_name ( required_db_cl_name , MYF ( 0 ) ) ;
2007-06-29 14:52:05 +02:00
if ( ! db_cl )
return 1 ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
fprintf ( sql_file ,
" ALTER DATABASE %s CHARACTER SET %s COLLATE %s %s \n " ,
2011-02-21 08:07:24 +01:00
( const char * ) quoted_db_name ,
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
( const char * ) db_cl - > csname ,
( const char * ) db_cl - > name ,
( const char * ) delimiter ) ;
2007-06-29 14:52:05 +02:00
* db_cl_altered = 1 ;
return 0 ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
}
2007-06-29 14:52:05 +02:00
* db_cl_altered = 0 ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
return 0 ;
}
2007-06-29 14:52:05 +02:00
static int restore_db_collation ( FILE * sql_file ,
const char * db_name ,
const char * delimiter ,
const char * db_cl_name )
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
{
2011-02-21 08:07:24 +01:00
char quoted_db_buf [ NAME_LEN * 2 + 3 ] ;
char * quoted_db_name = quote_name ( db_name , quoted_db_buf , FALSE ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
CHARSET_INFO * db_cl = get_charset_by_name ( db_cl_name , MYF ( 0 ) ) ;
2007-06-29 14:52:05 +02:00
if ( ! db_cl )
return 1 ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
fprintf ( sql_file ,
" ALTER DATABASE %s CHARACTER SET %s COLLATE %s %s \n " ,
2011-02-21 08:07:24 +01:00
( const char * ) quoted_db_name ,
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
( const char * ) db_cl - > csname ,
( const char * ) db_cl - > name ,
( const char * ) delimiter ) ;
2007-06-29 14:52:05 +02:00
return 0 ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
}
static void switch_cs_variables ( FILE * sql_file ,
const char * delimiter ,
const char * character_set_client ,
const char * character_set_results ,
const char * collation_connection )
{
fprintf ( sql_file ,
" /*!50003 SET @saved_cs_client = @@character_set_client */ %s \n "
" /*!50003 SET @saved_cs_results = @@character_set_results */ %s \n "
" /*!50003 SET @saved_col_connection = @@collation_connection */ %s \n "
" /*!50003 SET character_set_client = %s */ %s \n "
" /*!50003 SET character_set_results = %s */ %s \n "
" /*!50003 SET collation_connection = %s */ %s \n " ,
( const char * ) delimiter ,
( const char * ) delimiter ,
( const char * ) delimiter ,
( const char * ) character_set_client ,
( const char * ) delimiter ,
( const char * ) character_set_results ,
( const char * ) delimiter ,
( const char * ) collation_connection ,
( const char * ) delimiter ) ;
}
static void restore_cs_variables ( FILE * sql_file ,
const char * delimiter )
{
fprintf ( sql_file ,
" /*!50003 SET character_set_client = @saved_cs_client */ %s \n "
" /*!50003 SET character_set_results = @saved_cs_results */ %s \n "
" /*!50003 SET collation_connection = @saved_col_connection */ %s \n " ,
( const char * ) delimiter ,
( const char * ) delimiter ,
( const char * ) delimiter ) ;
}
static void switch_sql_mode ( FILE * sql_file ,
const char * delimiter ,
const char * sql_mode )
{
fprintf ( sql_file ,
" /*!50003 SET @saved_sql_mode = @@sql_mode */ %s \n "
" /*!50003 SET sql_mode = '%s' */ %s \n " ,
( const char * ) delimiter ,
( const char * ) sql_mode ,
( const char * ) delimiter ) ;
}
static void restore_sql_mode ( FILE * sql_file ,
const char * delimiter )
{
fprintf ( sql_file ,
" /*!50003 SET sql_mode = @saved_sql_mode */ %s \n " ,
( const char * ) delimiter ) ;
}
static void switch_time_zone ( FILE * sql_file ,
const char * delimiter ,
const char * time_zone )
{
fprintf ( sql_file ,
" /*!50003 SET @saved_time_zone = @@time_zone */ %s \n "
" /*!50003 SET time_zone = '%s' */ %s \n " ,
( const char * ) delimiter ,
( const char * ) time_zone ,
( const char * ) delimiter ) ;
}
static void restore_time_zone ( FILE * sql_file ,
const char * delimiter )
{
fprintf ( sql_file ,
" /*!50003 SET time_zone = @saved_time_zone */ %s \n " ,
( const char * ) delimiter ) ;
}
2007-07-12 10:49:39 +02:00
2008-03-12 22:03:50 +01:00
/**
Switch charset for results to some specified charset . If the server does not
support character_set_results variable , nothing can be done here . As for
whether something should be done here , future new callers of this function
should be aware that the server lacking the facility of switching charsets is
treated as success .
@ note If the server lacks support , then nothing is changed and no error
condition is returned .
@ returns whether there was an error or not
*/
2007-07-12 10:49:39 +02:00
static int switch_character_set_results ( MYSQL * mysql , const char * cs_name )
{
char query_buffer [ QUERY_LENGTH ] ;
size_t query_length ;
2008-03-12 22:03:50 +01:00
/* Server lacks facility. This is not an error, by arbitrary decision . */
if ( ! server_supports_switching_charsets )
return FALSE ;
2007-07-12 10:49:39 +02:00
query_length = my_snprintf ( query_buffer ,
sizeof ( query_buffer ) ,
" SET SESSION character_set_results = '%s' " ,
( const char * ) cs_name ) ;
return mysql_real_query ( mysql , query_buffer , query_length ) ;
}
2007-08-03 09:50:17 +02:00
/**
2009-10-07 22:57:03 +02:00
Rewrite statement , enclosing DEFINER clause in version - specific comment .
2007-08-03 09:50:17 +02:00
2009-10-07 22:57:03 +02:00
This function parses any CREATE statement and encloses DEFINER - clause in
version - specific comment :
2007-08-03 09:50:17 +02:00
input query : CREATE DEFINER = a @ b FUNCTION . . .
rewritten query : CREATE * / / * ! 50020 DEFINER = a @ b * / / * ! 50003 FUNCTION . . .
@ note This function will go away when WL # 3995 is implemented .
2009-10-07 22:57:03 +02:00
@ param [ in ] stmt_str CREATE statement string .
@ param [ in ] stmt_length Length of the stmt_str .
@ param [ in ] definer_version_str Minimal MySQL version number when
DEFINER clause is supported in the
given statement .
@ param [ in ] definer_version_length Length of definer_version_str .
@ param [ in ] stmt_version_str Minimal MySQL version number when the
given statement is supported .
@ param [ in ] stmt_version_length Length of stmt_version_str .
@ param [ in ] keyword_str Keyword to look for after CREATE .
@ param [ in ] keyword_length Length of keyword_str .
2007-08-03 09:50:17 +02:00
@ return pointer to the new allocated query string .
*/
2009-10-07 22:57:03 +02:00
static char * cover_definer_clause ( const char * stmt_str ,
uint stmt_length ,
const char * definer_version_str ,
uint definer_version_length ,
const char * stmt_version_str ,
uint stmt_version_length ,
const char * keyword_str ,
uint keyword_length )
2007-08-03 09:50:17 +02:00
{
2009-10-07 22:57:03 +02:00
char * definer_begin = my_case_str ( stmt_str , stmt_length ,
2007-08-03 09:50:17 +02:00
C_STRING_WITH_LEN ( " DEFINER " ) ) ;
2009-10-07 22:57:03 +02:00
char * definer_end = NULL ;
char * query_str = NULL ;
char * query_ptr ;
2007-08-03 09:50:17 +02:00
if ( ! definer_begin )
return NULL ;
definer_end = my_case_str ( definer_begin , strlen ( definer_begin ) ,
2009-10-07 22:57:03 +02:00
keyword_str , keyword_length ) ;
2007-08-03 09:50:17 +02:00
if ( ! definer_end )
2009-10-07 22:57:03 +02:00
return NULL ;
2007-08-03 09:50:17 +02:00
2009-10-07 22:57:03 +02:00
/*
Allocate memory for new query string : original string
from SHOW statement and version - specific comments .
*/
query_str = alloc_query_str ( stmt_length + 23 ) ;
2007-08-03 09:50:17 +02:00
2009-10-07 22:57:03 +02:00
query_ptr = strnmov ( query_str , stmt_str , definer_begin - stmt_str ) ;
query_ptr = strnmov ( query_ptr , C_STRING_WITH_LEN ( " */ /*! " ) ) ;
query_ptr = strnmov ( query_ptr , definer_version_str , definer_version_length ) ;
query_ptr = strnmov ( query_ptr , definer_begin , definer_end - definer_begin ) ;
query_ptr = strnmov ( query_ptr , C_STRING_WITH_LEN ( " */ /*! " ) ) ;
query_ptr = strnmov ( query_ptr , stmt_version_str , stmt_version_length ) ;
query_ptr = strxmov ( query_ptr , definer_end , NullS ) ;
2007-08-03 09:50:17 +02:00
return query_str ;
}
2006-02-21 13:21:17 +01:00
/*
Open a new . sql file to dump the table or view into
SYNOPSIS
open_sql_file_for_table
name name of the table or view
2009-05-27 15:16:21 +02:00
flags flags ( as per " man 2 open " )
2006-02-21 13:21:17 +01:00
RETURN VALUES
0 Failed to open file
> 0 Handle of the open file
*/
2009-05-27 15:16:21 +02:00
static FILE * open_sql_file_for_table ( const char * table , int flags )
2006-02-21 13:21:17 +01:00
{
FILE * res ;
char filename [ FN_REFLEN ] , tmp_path [ FN_REFLEN ] ;
convert_dirname ( tmp_path , path , NullS ) ;
res = my_fopen ( fn_format ( filename , table , tmp_path , " .sql " , 4 ) ,
2009-05-27 15:16:21 +02:00
flags , MYF ( MY_WME ) ) ;
2006-02-21 13:21:17 +01:00
return res ;
}
2004-11-10 17:56:45 +01:00
2006-11-20 21:42:06 +01:00
static void free_resources ( )
{
if ( md_result_file & & md_result_file ! = stdout )
my_fclose ( md_result_file , MYF ( 0 ) ) ;
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( opt_password ) ;
2009-10-14 18:37:38 +02:00
if ( my_hash_inited ( & ignore_table ) )
my_hash_free ( & ignore_table ) ;
2006-11-20 21:42:06 +01:00
if ( extended_insert )
dynstr_free ( & extended_row ) ;
if ( insert_pat_inited )
dynstr_free ( & insert_pat ) ;
if ( defaults_argv )
free_defaults ( defaults_argv ) ;
2007-08-01 21:59:05 +02:00
my_end ( my_end_arg ) ;
2006-11-20 21:42:06 +01:00
}
2007-03-22 18:35:29 +01:00
static void maybe_exit ( int error )
2000-07-31 21:29:14 +02:00
{
if ( ! first_error )
first_error = error ;
if ( ignore_errors )
return ;
2006-07-24 12:00:44 +02:00
if ( mysql )
mysql_close ( mysql ) ;
2006-11-20 21:42:06 +01:00
free_resources ( ) ;
2000-07-31 21:29:14 +02:00
exit ( error ) ;
}
/*
2006-11-20 21:42:06 +01:00
db_connect - - connects to the host and selects DB .
2000-07-31 21:29:14 +02:00
*/
2006-11-20 21:42:06 +01:00
static int connect_to_db ( char * host , char * user , char * passwd )
2000-07-31 21:29:14 +02:00
{
2004-03-11 15:46:27 +01:00
char buff [ 20 + FN_REFLEN ] ;
2006-11-20 21:42:06 +01:00
DBUG_ENTER ( " connect_to_db " ) ;
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Connecting to %s... \n " , host ? host : " localhost " ) ;
2000-07-31 21:29:14 +02:00
mysql_init ( & mysql_connection ) ;
if ( opt_compress )
mysql_options ( & mysql_connection , MYSQL_OPT_COMPRESS , NullS ) ;
# ifdef HAVE_OPENSSL
if ( opt_use_ssl )
mysql_ssl_set ( & mysql_connection , opt_ssl_key , opt_ssl_cert , opt_ssl_ca ,
2006-05-25 10:00:28 +02:00
opt_ssl_capath , opt_ssl_cipher ) ;
2006-04-18 17:58:27 +02:00
mysql_options ( & mysql_connection , MYSQL_OPT_SSL_VERIFY_SERVER_CERT ,
( char * ) & opt_ssl_verify_server_cert ) ;
2002-11-14 20:16:30 +01:00
# endif
if ( opt_protocol )
mysql_options ( & mysql_connection , MYSQL_OPT_PROTOCOL , ( char * ) & opt_protocol ) ;
# ifdef HAVE_SMEM
if ( shared_memory_base_name )
mysql_options ( & mysql_connection , MYSQL_SHARED_MEMORY_BASE_NAME , shared_memory_base_name ) ;
2000-07-31 21:29:14 +02:00
# endif
2004-05-01 11:07:42 +02:00
mysql_options ( & mysql_connection , MYSQL_SET_CHARSET_NAME , default_charset ) ;
2010-12-07 13:07:07 +01:00
if ( opt_plugin_dir & & * opt_plugin_dir )
mysql_options ( & mysql_connection , MYSQL_PLUGIN_DIR , opt_plugin_dir ) ;
if ( opt_default_auth & & * opt_default_auth )
mysql_options ( & mysql_connection , MYSQL_DEFAULT_AUTH , opt_default_auth ) ;
2006-07-24 12:00:44 +02:00
if ( ! ( mysql = mysql_real_connect ( & mysql_connection , host , user , passwd ,
2006-11-20 21:42:06 +01:00
NULL , opt_mysql_port , opt_mysql_unix_port ,
0 ) ) )
2000-07-31 21:29:14 +02:00
{
2005-04-04 23:32:48 +02:00
DB_error ( & mysql_connection , " when trying to connect " ) ;
2006-11-20 21:42:06 +01:00
DBUG_RETURN ( 1 ) ;
2000-07-31 21:29:14 +02:00
}
2009-02-02 18:19:07 +01:00
if ( ( mysql_get_server_version ( & mysql_connection ) < 40100 ) | |
( opt_compatible_mode & 3 ) )
2008-03-12 22:03:50 +01:00
{
/* Don't dump SET NAMES with a pre-4.1 server (bug#7997). */
2005-02-25 15:12:53 +01:00
opt_set_charset = 0 ;
2008-03-12 22:03:50 +01:00
/* Don't switch charsets for 4.1 and earlier. (bug#34192). */
server_supports_switching_charsets = FALSE ;
}
2004-11-10 17:56:45 +01:00
/*
As we ' re going to set SQL_MODE , it would be lost on reconnect , so we
cannot reconnect .
*/
2006-07-24 12:00:44 +02:00
mysql - > reconnect = 0 ;
2005-03-16 21:56:56 +01:00
my_snprintf ( buff , sizeof ( buff ) , " /*!40100 SET @@SQL_MODE='%s' */ " ,
2006-05-25 10:00:28 +02:00
compatible_mode_normal_str ) ;
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , 0 , buff ) )
2006-11-20 21:42:06 +01:00
DBUG_RETURN ( 1 ) ;
2005-10-13 07:44:42 +02:00
/*
2006-05-25 10:00:28 +02:00
set time_zone to UTC to allow dumping date types between servers with
2005-10-13 07:44:42 +02:00
different time zone settings
*/
if ( opt_tz_utc )
{
my_snprintf ( buff , sizeof ( buff ) , " /*!40103 SET TIME_ZONE='+00:00' */ " ) ;
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , 0 , buff ) )
2006-11-20 21:42:06 +01:00
DBUG_RETURN ( 1 ) ;
2005-10-13 07:44:42 +02:00
}
2006-11-20 21:42:06 +01:00
DBUG_RETURN ( 0 ) ;
} /* connect_to_db */
2000-07-31 21:29:14 +02:00
/*
* * dbDisconnect - - disconnects from the host .
*/
static void dbDisconnect ( char * host )
{
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Disconnecting from %s... \n " , host ? host : " localhost " ) ;
2006-07-24 12:00:44 +02:00
mysql_close ( mysql ) ;
2000-07-31 21:29:14 +02:00
} /* dbDisconnect */
static void unescape ( FILE * file , char * pos , uint length )
{
char * tmp ;
DBUG_ENTER ( " unescape " ) ;
if ( ! ( tmp = ( char * ) my_malloc ( length * 2 + 1 , MYF ( MY_WME ) ) ) )
2007-03-22 18:35:29 +01:00
die ( EX_MYSQLERR , " Couldn't allocate memory " ) ;
2001-11-05 22:48:03 +01:00
mysql_real_escape_string ( & mysql_connection , tmp , pos , length ) ;
2000-07-31 21:29:14 +02:00
fputc ( ' \' ' , file ) ;
fputs ( tmp , file ) ;
fputc ( ' \' ' , file ) ;
2004-06-24 22:14:12 +02:00
check_io ( file ) ;
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( tmp ) ;
2000-07-31 21:29:14 +02:00
DBUG_VOID_RETURN ;
} /* unescape */
static my_bool test_if_special_chars ( const char * str )
{
# if MYSQL_VERSION_ID >= 32300
for ( ; * str ; str + + )
2003-03-16 08:20:45 +01:00
if ( ! my_isvar ( charset_info , * str ) & & * str ! = ' $ ' )
2000-07-31 21:29:14 +02:00
return 1 ;
# endif
return 0 ;
} /* test_if_special_chars */
2003-09-03 17:48:10 +02:00
2003-11-04 08:40:36 +01:00
2005-10-13 22:42:56 +02:00
/*
quote_name ( name , buff , force )
2006-05-25 10:00:28 +02:00
Quotes char string , taking into account compatible mode
2005-10-13 22:42:56 +02:00
Args
name Unquoted string containing that which will be quoted
buff The buffer that contains the quoted value , also returned
2006-05-25 10:00:28 +02:00
force Flag to make it ignore ' test_if_special_chars '
2005-10-13 22:42:56 +02:00
Returns
buff quoted string
*/
2003-09-03 17:48:10 +02:00
static char * quote_name ( const char * name , char * buff , my_bool force )
2000-07-31 21:29:14 +02:00
{
2003-09-03 17:48:10 +02:00
char * to = buff ;
2004-03-11 15:46:27 +01:00
char qtype = ( opt_compatible_mode & MASK_ANSI_QUOTES ) ? ' \" ' : ' ` ' ;
2003-09-03 17:48:10 +02:00
if ( ! force & & ! opt_quoted & & ! test_if_special_chars ( name ) )
return ( char * ) name ;
2004-03-11 15:46:27 +01:00
* to + + = qtype ;
2003-09-03 17:48:10 +02:00
while ( * name )
{
2004-03-11 15:46:27 +01:00
if ( * name = = qtype )
* to + + = qtype ;
2003-09-03 17:48:10 +02:00
* to + + = * name + + ;
}
2004-03-11 15:46:27 +01:00
to [ 0 ] = qtype ;
to [ 1 ] = 0 ;
2000-07-31 21:29:14 +02:00
return buff ;
} /* quote_name */
2005-06-27 19:31:00 +02:00
2005-06-21 14:19:56 +02:00
/*
Quote a table name so it can be used in " SHOW TABLES LIKE <tabname> "
2000-07-31 21:29:14 +02:00
2005-06-21 14:19:56 +02:00
SYNOPSIS
2005-06-27 19:31:00 +02:00
quote_for_like ( )
name name of the table
buff quoted name of the table
2005-06-21 14:19:56 +02:00
DESCRIPTION
Quote \ , _ , ' and % characters
2000-07-31 21:29:14 +02:00
2005-06-21 14:19:56 +02:00
Note : Because MySQL uses the C escape syntax in strings
( for example , ' \n ' to represent newline ) , you must double
any ' \ ' that you use in your LIKE strings . For example , to
search for ' \n ' , specify it as ' \ \ n ' . To search for ' \ ' , specify
it as ' \ \ \ \ ' ( the backslashes are stripped once by the parser
and another time when the pattern match is done , leaving a
single backslash to be matched ) .
Example : " t \1 " = > " t \\ \\ 1 "
*/
2003-12-14 05:39:52 +01:00
static char * quote_for_like ( const char * name , char * buff )
{
char * to = buff ;
* to + + = ' \' ' ;
while ( * name )
{
2005-06-21 14:19:56 +02:00
if ( * name = = ' \\ ' )
{
* to + + = ' \\ ' ;
* to + + = ' \\ ' ;
* to + + = ' \\ ' ;
}
else if ( * name = = ' \' ' | | * name = = ' _ ' | | * name = = ' % ' )
2003-12-14 05:39:52 +01:00
* to + + = ' \\ ' ;
* to + + = * name + + ;
}
to [ 0 ] = ' \' ' ;
to [ 1 ] = 0 ;
return buff ;
}
2012-01-10 09:03:45 +01:00
/**
2003-11-03 13:49:39 +01:00
Quote and print a string .
2006-05-25 10:00:28 +02:00
2012-01-10 09:03:45 +01:00
@ param xml_file - Output file .
@ param str - String to print .
@ param len - Its length .
@ param is_attribute_name - A check for attribute name or value .
2006-05-25 10:00:28 +02:00
2012-01-10 09:03:45 +01:00
@ description
2004-11-16 23:45:24 +01:00
Quote ' < ' ' > ' ' & ' ' \" ' chars and print a string to the xml_file .
2003-11-03 13:49:39 +01:00
*/
2012-01-10 09:03:45 +01:00
static void print_quoted_xml ( FILE * xml_file , const char * str , ulong len ,
my_bool is_attribute_name )
2003-11-03 13:49:39 +01:00
{
const char * end ;
2006-05-25 10:00:28 +02:00
2003-11-03 13:49:39 +01:00
for ( end = str + len ; str ! = end ; str + + )
{
switch ( * str ) {
case ' < ' :
fputs ( " < " , xml_file ) ;
break ;
case ' > ' :
fputs ( " > " , xml_file ) ;
break ;
case ' & ' :
fputs ( " & " , xml_file ) ;
break ;
case ' \" ' :
fputs ( " " " , xml_file ) ;
break ;
2012-01-10 09:03:45 +01:00
case ' ' :
/* Attribute names cannot contain spaces. */
if ( is_attribute_name )
{
fputs ( " _ " , xml_file ) ;
break ;
}
/* fall through */
2003-11-03 13:49:39 +01:00
default :
fputc ( * str , xml_file ) ;
break ;
}
}
2004-06-24 22:14:12 +02:00
check_io ( xml_file ) ;
2003-11-03 13:49:39 +01:00
}
/*
2006-10-19 00:43:51 +02:00
Print xml tag . Optionally add attribute ( s ) .
2006-05-25 10:00:28 +02:00
2003-11-03 13:49:39 +01:00
SYNOPSIS
2006-10-19 00:43:51 +02:00
print_xml_tag ( xml_file , sbeg , send , tag_name , first_attribute_name ,
. . . , attribute_name_n , attribute_value_n , NullS )
xml_file - output file
sbeg - line beginning
2006-12-14 23:51:37 +01:00
line_end - line ending
2006-10-19 00:43:51 +02:00
tag_name - XML tag name .
first_attribute_name - tag and first attribute
first_attribute_value - ( Implied ) value of first attribute
attribute_name_n - attribute n
attribute_value_n - value of attribute n
2006-05-25 10:00:28 +02:00
2003-11-03 13:49:39 +01:00
DESCRIPTION
2006-10-19 00:43:51 +02:00
Print XML tag with any number of attribute = " value " pairs to the xml_file .
Format is :
sbeg < tag_name first_attribute_name = " first_attribute_value " . . .
attribute_name_n = " attribute_value_n " > send
2003-11-03 13:49:39 +01:00
NOTE
2006-10-19 00:43:51 +02:00
Additional arguments must be present in attribute / value pairs .
The last argument should be the null character pointer .
All attribute_value arguments MUST be NULL terminated strings .
All attribute_value arguments will be quoted before output .
2003-11-03 13:49:39 +01:00
*/
2006-12-14 23:51:37 +01:00
static void print_xml_tag ( FILE * xml_file , const char * sbeg ,
const char * line_end ,
2006-10-19 00:43:51 +02:00
const char * tag_name ,
const char * first_attribute_name , . . . )
2003-11-03 13:49:39 +01:00
{
2006-10-19 00:43:51 +02:00
va_list arg_list ;
2006-10-24 13:52:50 +02:00
const char * attribute_name , * attribute_value ;
2006-10-19 00:43:51 +02:00
2003-11-03 13:49:39 +01:00
fputs ( sbeg , xml_file ) ;
2006-10-19 00:43:51 +02:00
fputc ( ' < ' , xml_file ) ;
fputs ( tag_name , xml_file ) ;
va_start ( arg_list , first_attribute_name ) ;
attribute_name = first_attribute_name ;
while ( attribute_name ! = NullS )
{
attribute_value = va_arg ( arg_list , char * ) ;
DBUG_ASSERT ( attribute_value ! = NullS ) ;
fputc ( ' ' , xml_file ) ;
fputs ( attribute_name , xml_file ) ;
fputc ( ' \" ' , xml_file ) ;
2012-01-10 09:03:45 +01:00
print_quoted_xml ( xml_file , attribute_value , strlen ( attribute_value ) , 0 ) ;
2006-10-19 00:43:51 +02:00
fputc ( ' \" ' , xml_file ) ;
attribute_name = va_arg ( arg_list , char * ) ;
}
va_end ( arg_list ) ;
fputc ( ' > ' , xml_file ) ;
2006-12-14 23:51:37 +01:00
fputs ( line_end , xml_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( xml_file ) ;
2003-11-03 13:49:39 +01:00
}
2005-06-22 20:22:54 +02:00
/*
Print xml tag with for a field that is null
SYNOPSIS
print_xml_null_tag ( )
2006-05-25 10:00:28 +02:00
xml_file - output file
sbeg - line beginning
stag_atr - tag and attribute
sval - value of attribute
2006-12-14 23:51:37 +01:00
line_end - line ending
2005-06-22 20:22:54 +02:00
DESCRIPTION
Print tag with one attribute to the xml_file . Format is :
< stag_atr = " sval " xsi : nil = " true " / >
NOTE
sval MUST be a NULL terminated string .
sval string will be qouted before output .
*/
static void print_xml_null_tag ( FILE * xml_file , const char * sbeg ,
const char * stag_atr , const char * sval ,
2006-12-14 23:51:37 +01:00
const char * line_end )
2005-06-22 20:22:54 +02:00
{
fputs ( sbeg , xml_file ) ;
fputs ( " < " , xml_file ) ;
fputs ( stag_atr , xml_file ) ;
fputs ( " \" " , xml_file ) ;
2012-01-10 09:03:45 +01:00
print_quoted_xml ( xml_file , sval , strlen ( sval ) , 0 ) ;
2005-06-22 20:22:54 +02:00
fputs ( " \" xsi:nil= \" true \" /> " , xml_file ) ;
2006-12-14 23:51:37 +01:00
fputs ( line_end , xml_file ) ;
2005-06-22 20:22:54 +02:00
check_io ( xml_file ) ;
}
2012-01-10 09:03:45 +01:00
/**
Print xml CDATA section .
@ param xml_file - output file
@ param str - string to print
@ param len - length of the string
@ note
This function also takes care of the presence of ' [ [ > '
string in the str . If found , the CDATA section is broken
into two CDATA sections , < ! [ CDATA [ ] ] ] ] > and < ! [ CDATA [ > ] ] .
*/
static void print_xml_cdata ( FILE * xml_file , const char * str , ulong len )
{
const char * end ;
fputs ( " <![CDATA[ \n " , xml_file ) ;
for ( end = str + len ; str ! = end ; str + + )
{
switch ( * str ) {
case ' ] ' :
if ( ( * ( str + 1 ) = = ' ] ' ) & & ( * ( str + 2 ) = = ' > ' ) )
{
fputs ( " ]]]]><![CDATA[> " , xml_file ) ;
str + = 2 ;
continue ;
}
/* fall through */
default :
fputc ( * str , xml_file ) ;
break ;
}
}
fputs ( " \n ]]> \n " , xml_file ) ;
check_io ( xml_file ) ;
}
2003-11-03 13:49:39 +01:00
/*
Print xml tag with many attributes .
SYNOPSIS
print_xml_row ( )
2006-05-25 10:00:28 +02:00
xml_file - output file
row_name - xml tag name
tableRes - query result
row - result row
2012-01-10 09:03:45 +01:00
str_create - create statement header string
2006-05-25 10:00:28 +02:00
2003-11-03 13:49:39 +01:00
DESCRIPTION
Print tag with many attribute to the xml_file . Format is :
\ t \ t < row_name Atr1 = " Val1 " Atr2 = " Val2 " . . . / >
NOTE
All atributes and values will be quoted before output .
*/
static void print_xml_row ( FILE * xml_file , const char * row_name ,
2012-01-10 09:03:45 +01:00
MYSQL_RES * tableRes , MYSQL_ROW * row ,
const char * str_create )
2003-10-30 11:58:30 +01:00
{
uint i ;
2012-01-10 09:03:45 +01:00
my_bool body_found = 0 ;
2012-01-12 16:27:53 +01:00
char * create_stmt_ptr = NULL ;
2012-01-10 09:03:45 +01:00
ulong create_stmt_len = 0 ;
2003-10-30 11:58:30 +01:00
MYSQL_FIELD * field ;
ulong * lengths = mysql_fetch_lengths ( tableRes ) ;
2005-09-04 01:34:58 +02:00
2003-10-30 11:58:30 +01:00
fprintf ( xml_file , " \t \t <%s " , row_name ) ;
2004-06-24 22:14:12 +02:00
check_io ( xml_file ) ;
2003-10-30 11:58:30 +01:00
mysql_field_seek ( tableRes , 0 ) ;
for ( i = 0 ; ( field = mysql_fetch_field ( tableRes ) ) ; i + + )
{
2003-11-03 13:49:39 +01:00
if ( ( * row ) [ i ] )
2003-10-30 11:58:30 +01:00
{
2012-01-10 09:03:45 +01:00
/* For 'create' statements, dump using CDATA. */
if ( ( str_create ) & & ( strcmp ( str_create , field - > name ) = = 0 ) )
{
create_stmt_ptr = ( * row ) [ i ] ;
create_stmt_len = lengths [ i ] ;
body_found = 1 ;
}
else
{
fputc ( ' ' , xml_file ) ;
print_quoted_xml ( xml_file , field - > name , field - > name_length , 1 ) ;
fputs ( " = \" " , xml_file ) ;
print_quoted_xml ( xml_file , ( * row ) [ i ] , lengths [ i ] , 0 ) ;
fputc ( ' " ' , xml_file ) ;
check_io ( xml_file ) ;
}
2003-10-30 11:58:30 +01:00
}
}
2012-01-10 09:03:45 +01:00
if ( create_stmt_len )
{
DBUG_ASSERT ( body_found ) ;
fputs ( " > \n " , xml_file ) ;
print_xml_cdata ( xml_file , create_stmt_ptr , create_stmt_len ) ;
fprintf ( xml_file , " \t \t </%s> \n " , row_name ) ;
}
else
fputs ( " /> \n " , xml_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( xml_file ) ;
2003-10-30 11:58:30 +01:00
}
2006-03-09 21:12:43 +01:00
2012-01-10 09:03:45 +01:00
/**
Print xml comments .
@ param xml_file - output file
@ param len - length of comment message
@ param comment_string - comment message
@ description
Print the comment message in the format :
" <!-- \n comment string \n --> \n "
@ note
Any occurrence of continuous hyphens will be
squeezed to a single hyphen .
*/
static void print_xml_comment ( FILE * xml_file , ulong len ,
const char * comment_string )
{
const char * end ;
fputs ( " <!-- " , xml_file ) ;
for ( end = comment_string + len ; comment_string ! = end ; comment_string + + )
{
/*
The string " -- " ( double - hyphen ) MUST NOT occur within xml comments .
*/
switch ( * comment_string ) {
case ' - ' :
if ( * ( comment_string + 1 ) = = ' - ' ) /* Only one hyphen allowed. */
break ;
default :
fputc ( * comment_string , xml_file ) ;
break ;
2003-10-30 11:58:30 +01:00
}
}
2012-01-10 09:03:45 +01:00
fputs ( " --> \n " , xml_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( xml_file ) ;
2003-10-30 11:58:30 +01:00
}
2006-03-09 21:12:43 +01:00
2012-01-10 09:03:45 +01:00
/* A common printing function for xml and non-xml modes. */
static void print_comment ( FILE * sql_file , my_bool is_error , const char * format ,
. . . )
{
static char comment_buff [ COMMENT_LENGTH ] ;
va_list args ;
/* If its an error message, print it ignoring opt_comments. */
if ( ! is_error & & ! opt_comments )
return ;
va_start ( args , format ) ;
my_vsnprintf ( comment_buff , COMMENT_LENGTH , format , args ) ;
va_end ( args ) ;
if ( ! opt_xml )
{
fputs ( comment_buff , sql_file ) ;
check_io ( sql_file ) ;
return ;
}
print_xml_comment ( sql_file , strlen ( comment_buff ) , comment_buff ) ;
}
2006-03-09 21:12:43 +01:00
/*
create_delimiter
Generate a new ( null - terminated ) string that does not exist in query
and is therefore suitable for use as a query delimiter . Store this
delimiter in delimiter_buff .
This is quite simple in that it doesn ' t even try to parse statements as an
interpreter would . It merely returns a string that is not in the query , which
is much more than adequate for constructing a delimiter .
RETURN
ptr to the delimiter on Success
NULL on Failure
*/
static char * create_delimiter ( char * query , char * delimiter_buff ,
int delimiter_max_size )
{
int proposed_length ;
char * presence ;
delimiter_buff [ 0 ] = ' ; ' ; /* start with one semicolon, and */
for ( proposed_length = 2 ; proposed_length < delimiter_max_size ;
delimiter_max_size + + ) {
delimiter_buff [ proposed_length - 1 ] = ' ; ' ; /* add semicolons, until */
delimiter_buff [ proposed_length ] = ' \0 ' ;
presence = strstr ( query , delimiter_buff ) ;
if ( presence = = NULL ) { /* the proposed delimiter is not in the query. */
return delimiter_buff ;
}
}
return NULL ; /* but if we run out of space, return nothing at all. */
}
/*
dump_events_for_db
- - retrieves list of events for a given db , and prints out
the CREATE EVENT statement into the output ( the dump ) .
RETURN
0 Success
1 Error
*/
static uint dump_events_for_db ( char * db )
{
char query_buff [ QUERY_LENGTH ] ;
char db_name_buff [ NAME_LEN * 2 + 3 ] , name_buff [ NAME_LEN * 2 + 3 ] ;
char * event_name ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
char delimiter [ QUERY_LENGTH ] ;
2006-03-09 21:12:43 +01:00
FILE * sql_file = md_result_file ;
MYSQL_RES * event_res , * event_list_res ;
MYSQL_ROW row , event_list_row ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
char db_cl_name [ MY_CS_NAME_SIZE ] ;
2007-08-06 20:52:50 +02:00
int db_cl_altered = FALSE ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2006-03-09 21:12:43 +01:00
DBUG_ENTER ( " dump_events_for_db " ) ;
DBUG_PRINT ( " enter " , ( " db: '%s' " , db ) ) ;
2006-08-14 11:27:11 +02:00
mysql_real_escape_string ( mysql , db_name_buff , db , strlen ( db ) ) ;
2006-03-09 21:12:43 +01:00
/* nice comments */
2012-01-10 09:03:45 +01:00
print_comment ( sql_file , 0 ,
" \n -- \n -- Dumping events for database '%s' \n -- \n " , db ) ;
2006-03-09 21:12:43 +01:00
/*
not using " mysql_query_with_error_report " because we may have not
enough privileges to lock mysql . events .
*/
if ( lock_tables )
2006-08-14 11:27:11 +02:00
mysql_query ( mysql , " LOCK TABLES mysql.event READ " ) ;
2006-03-09 21:12:43 +01:00
2006-08-14 11:27:11 +02:00
if ( mysql_query_with_error_report ( mysql , & event_list_res , " show events " ) )
2006-03-09 21:12:43 +01:00
DBUG_RETURN ( 0 ) ;
strcpy ( delimiter , " ; " ) ;
if ( mysql_num_rows ( event_list_res ) > 0 )
{
2012-01-10 09:03:45 +01:00
if ( opt_xml )
fputs ( " \t <events> \n " , sql_file ) ;
else
{
fprintf ( sql_file , " /*!50106 SET @save_time_zone= @@TIME_ZONE */ ; \n " ) ;
2007-03-16 15:31:07 +01:00
2012-01-10 09:03:45 +01:00
/* Get database collation. */
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2012-01-10 09:03:45 +01:00
if ( fetch_db_collation ( db_name_buff , db_cl_name , sizeof ( db_cl_name ) ) )
DBUG_RETURN ( 1 ) ;
}
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-07-12 10:49:39 +02:00
if ( switch_character_set_results ( mysql , " binary " ) )
DBUG_RETURN ( 1 ) ;
2006-03-09 21:12:43 +01:00
while ( ( event_list_row = mysql_fetch_row ( event_list_res ) ) ! = NULL )
{
event_name = quote_name ( event_list_row [ 1 ] , name_buff , 0 ) ;
DBUG_PRINT ( " info " , ( " retrieving CREATE EVENT for %s " , name_buff ) ) ;
my_snprintf ( query_buff , sizeof ( query_buff ) , " SHOW CREATE EVENT %s " ,
event_name ) ;
2006-08-14 11:27:11 +02:00
if ( mysql_query_with_error_report ( mysql , & event_res , query_buff ) )
2006-03-09 21:12:43 +01:00
DBUG_RETURN ( 1 ) ;
while ( ( row = mysql_fetch_row ( event_res ) ) ! = NULL )
{
2012-01-10 09:03:45 +01:00
if ( opt_xml )
{
print_xml_row ( sql_file , " event " , event_res , & row ,
" Create Event " ) ;
continue ;
}
2006-03-09 21:12:43 +01:00
/*
if the user has EXECUTE privilege he can see event names , but not the
event body !
*/
2007-03-16 15:31:07 +01:00
if ( strlen ( row [ 3 ] ) ! = 0 )
2006-03-09 21:12:43 +01:00
{
2009-10-07 22:57:03 +02:00
char * query_str ;
2006-03-09 21:12:43 +01:00
if ( opt_drop )
fprintf ( sql_file , " /*!50106 DROP EVENT IF EXISTS %s */%s \n " ,
event_name , delimiter ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
if ( create_delimiter ( row [ 3 ] , delimiter , sizeof ( delimiter ) ) = = NULL )
{
fprintf ( stderr , " %s: Warning: Can't create delimiter for event '%s' \n " ,
2007-10-08 23:43:32 +02:00
my_progname , event_name ) ;
2006-03-09 21:12:43 +01:00
DBUG_RETURN ( 1 ) ;
}
fprintf ( sql_file , " DELIMITER %s \n " , delimiter ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-08-03 09:50:17 +02:00
if ( mysql_num_fields ( event_res ) > = 7 )
2007-06-29 14:52:05 +02:00
{
2007-08-03 09:50:17 +02:00
if ( switch_db_collation ( sql_file , db_name_buff , delimiter ,
db_cl_name , row [ 6 ] , & db_cl_altered ) )
{
DBUG_RETURN ( 1 ) ;
}
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-08-03 09:50:17 +02:00
switch_cs_variables ( sql_file , delimiter ,
row [ 4 ] , /* character_set_client */
row [ 4 ] , /* character_set_results */
row [ 5 ] ) ; /* collation_connection */
}
2009-10-07 22:57:03 +02:00
else
{
/*
mysqldump is being run against the server , that does not
provide character set information in SHOW CREATE
statements .
2007-08-03 09:50:17 +02:00
2009-10-07 22:57:03 +02:00
NOTE : the dump may be incorrect , since character set
information is required in order to restore event properly .
*/
2007-08-03 09:50:17 +02:00
2009-10-07 22:57:03 +02:00
fprintf ( sql_file ,
" -- \n "
" -- WARNING: old server version. "
" The following dump may be incomplete. \n "
" -- \n " ) ;
}
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
switch_sql_mode ( sql_file , delimiter , row [ 1 ] ) ;
switch_time_zone ( sql_file , delimiter , row [ 2 ] ) ;
2009-10-07 22:57:03 +02:00
query_str = cover_definer_clause ( row [ 3 ] , strlen ( row [ 3 ] ) ,
C_STRING_WITH_LEN ( " 50117 " ) ,
C_STRING_WITH_LEN ( " 50106 " ) ,
C_STRING_WITH_LEN ( " EVENT " ) ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
fprintf ( sql_file ,
" /*!50106 %s */ %s \n " ,
2009-10-07 22:57:03 +02:00
( const char * ) ( query_str ! = NULL ? query_str : row [ 3 ] ) ,
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
( const char * ) delimiter ) ;
restore_time_zone ( sql_file , delimiter ) ;
restore_sql_mode ( sql_file , delimiter ) ;
2007-08-03 09:50:17 +02:00
if ( mysql_num_fields ( event_res ) > = 7 )
2007-06-29 14:52:05 +02:00
{
2007-08-03 09:50:17 +02:00
restore_cs_variables ( sql_file , delimiter ) ;
if ( db_cl_altered )
{
if ( restore_db_collation ( sql_file , db_name_buff , delimiter ,
db_cl_name ) )
DBUG_RETURN ( 1 ) ;
}
2007-06-29 14:52:05 +02:00
}
2006-03-09 21:12:43 +01:00
}
} /* end of event printing */
2007-03-24 00:30:25 +01:00
mysql_free_result ( event_res ) ;
2006-03-09 21:12:43 +01:00
} /* end of list of events */
2012-01-10 09:03:45 +01:00
if ( opt_xml )
{
fputs ( " \t </events> \n " , sql_file ) ;
check_io ( sql_file ) ;
}
else
{
fprintf ( sql_file , " DELIMITER ; \n " ) ;
fprintf ( sql_file , " /*!50106 SET TIME_ZONE= @save_time_zone */ ; \n " ) ;
}
2007-07-12 10:49:39 +02:00
if ( switch_character_set_results ( mysql , default_charset ) )
DBUG_RETURN ( 1 ) ;
2006-03-09 21:12:43 +01:00
}
mysql_free_result ( event_list_res ) ;
if ( lock_tables )
2009-11-24 14:54:59 +01:00
( void ) mysql_query_with_error_report ( mysql , 0 , " UNLOCK TABLES " ) ;
2006-03-09 21:12:43 +01:00
DBUG_RETURN ( 0 ) ;
}
2006-10-19 00:43:51 +02:00
/*
Print hex value for blob data .
SYNOPSIS
print_blob_as_hex ( )
output_file - output file
str - string to print
len - its length
DESCRIPTION
Print hex value for blob data .
*/
static void print_blob_as_hex ( FILE * output_file , const char * str , ulong len )
{
/* sakaik got the idea to to provide blob's in hex notation. */
2006-10-24 13:52:50 +02:00
const char * ptr = str , * end = ptr + len ;
2006-10-19 00:43:51 +02:00
for ( ; ptr < end ; ptr + + )
fprintf ( output_file , " %02X " , * ( ( uchar * ) ptr ) ) ;
check_io ( output_file ) ;
}
2005-09-04 01:34:58 +02:00
/*
dump_routines_for_db
2006-03-09 21:12:43 +01:00
- - retrieves list of routines for a given db , and prints out
2005-09-04 01:34:58 +02:00
the CREATE PROCEDURE definition into the output ( the dump ) .
This function has logic to print the appropriate syntax depending on whether
this is a procedure or functions
2005-10-06 16:54:43 +02:00
RETURN
0 Success
1 Error
2005-09-04 01:34:58 +02:00
*/
2005-10-06 16:54:43 +02:00
static uint dump_routines_for_db ( char * db )
2005-09-04 01:34:58 +02:00
{
2006-03-09 21:12:43 +01:00
char query_buff [ QUERY_LENGTH ] ;
2005-10-06 16:54:43 +02:00
const char * routine_type [ ] = { " FUNCTION " , " PROCEDURE " } ;
char db_name_buff [ NAME_LEN * 2 + 3 ] , name_buff [ NAME_LEN * 2 + 3 ] ;
char * routine_name ;
2005-09-04 01:34:58 +02:00
int i ;
2005-10-06 16:54:43 +02:00
FILE * sql_file = md_result_file ;
2005-09-13 19:32:37 +02:00
MYSQL_RES * routine_res , * routine_list_res ;
2005-09-08 19:45:25 +02:00
MYSQL_ROW row , routine_list_row ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
char db_cl_name [ MY_CS_NAME_SIZE ] ;
2007-08-06 20:52:50 +02:00
int db_cl_altered = FALSE ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2005-09-04 01:34:58 +02:00
DBUG_ENTER ( " dump_routines_for_db " ) ;
2005-10-06 16:54:43 +02:00
DBUG_PRINT ( " enter " , ( " db: '%s' " , db ) ) ;
2005-09-04 01:34:58 +02:00
2006-07-24 12:00:44 +02:00
mysql_real_escape_string ( mysql , db_name_buff , db , strlen ( db ) ) ;
2005-09-04 01:34:58 +02:00
/* nice comments */
2012-01-10 09:03:45 +01:00
print_comment ( sql_file , 0 ,
" \n -- \n -- Dumping routines for database '%s' \n -- \n " , db ) ;
2005-09-08 19:45:25 +02:00
/*
2005-09-13 19:32:37 +02:00
not using " mysql_query_with_error_report " because we may have not
enough privileges to lock mysql . proc .
2005-09-08 19:45:25 +02:00
*/
2005-09-13 19:32:37 +02:00
if ( lock_tables )
2006-07-24 12:00:44 +02:00
mysql_query ( mysql , " LOCK TABLES mysql.proc READ " ) ;
2005-09-08 19:45:25 +02:00
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
/* Get database collation. */
2007-05-14 09:02:40 +02:00
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
if ( fetch_db_collation ( db_name_buff , db_cl_name , sizeof ( db_cl_name ) ) )
DBUG_RETURN ( 1 ) ;
2005-09-04 01:34:58 +02:00
2007-07-12 10:49:39 +02:00
if ( switch_character_set_results ( mysql , " binary " ) )
DBUG_RETURN ( 1 ) ;
2012-01-10 09:03:45 +01:00
if ( opt_xml )
fputs ( " \t <routines> \n " , sql_file ) ;
2005-09-04 01:34:58 +02:00
/* 0, retrieve and dump functions, 1, procedures */
2005-10-06 16:54:43 +02:00
for ( i = 0 ; i < = 1 ; i + + )
2005-09-04 01:34:58 +02:00
{
my_snprintf ( query_buff , sizeof ( query_buff ) ,
" SHOW %s STATUS WHERE Db = '%s' " ,
2005-09-13 19:32:37 +02:00
routine_type [ i ] , db_name_buff ) ;
2005-09-04 01:34:58 +02:00
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & routine_list_res , query_buff ) )
2005-09-04 01:34:58 +02:00
DBUG_RETURN ( 1 ) ;
if ( mysql_num_rows ( routine_list_res ) )
{
2005-10-06 16:54:43 +02:00
while ( ( routine_list_row = mysql_fetch_row ( routine_list_res ) ) )
2005-09-04 01:34:58 +02:00
{
2006-03-09 21:12:43 +01:00
routine_name = quote_name ( routine_list_row [ 1 ] , name_buff , 0 ) ;
2005-09-13 19:32:37 +02:00
DBUG_PRINT ( " info " , ( " retrieving CREATE %s for %s " , routine_type [ i ] ,
name_buff ) ) ;
2005-09-04 01:34:58 +02:00
my_snprintf ( query_buff , sizeof ( query_buff ) , " SHOW CREATE %s %s " ,
2005-09-13 19:32:37 +02:00
routine_type [ i ] , routine_name ) ;
2005-09-04 01:34:58 +02:00
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & routine_res , query_buff ) )
2005-09-04 01:34:58 +02:00
DBUG_RETURN ( 1 ) ;
2005-10-06 16:54:43 +02:00
while ( ( row = mysql_fetch_row ( routine_res ) ) )
2005-09-04 01:34:58 +02:00
{
/*
2005-09-13 19:32:37 +02:00
if the user has EXECUTE privilege he see routine names , but NOT the
routine body of other routines that are not the creator of !
2005-09-04 01:34:58 +02:00
*/
2007-04-30 12:32:57 +02:00
DBUG_PRINT ( " info " , ( " length of body for %s row[2] '%s' is %d " ,
routine_name , row [ 2 ] ? row [ 2 ] : " (null) " ,
row [ 2 ] ? ( int ) strlen ( row [ 2 ] ) : 0 ) ) ;
if ( row [ 2 ] = = NULL )
{
2012-01-10 09:03:45 +01:00
print_comment ( sql_file , 1 , " \n -- insufficient privileges to %s \n " ,
query_buff ) ;
print_comment ( sql_file , 1 ,
" -- does %s have permissions on mysql.proc? \n \n " ,
current_user ) ;
2007-04-30 12:32:57 +02:00
maybe_die ( EX_MYSQLERR , " %s has insufficent privileges to %s! " , current_user , query_buff ) ;
}
else if ( strlen ( row [ 2 ] ) )
2005-09-04 01:34:58 +02:00
{
2012-01-10 09:03:45 +01:00
if ( opt_xml )
{
2012-02-28 17:20:30 +01:00
if ( i ) /* Procedures. */
2012-01-10 09:03:45 +01:00
print_xml_row ( sql_file , " routine " , routine_res , & row ,
" Create Procedure " ) ;
2012-02-28 17:20:30 +01:00
else /* Functions. */
2012-01-10 09:03:45 +01:00
print_xml_row ( sql_file , " routine " , routine_res , & row ,
" Create Function " ) ;
continue ;
}
2005-09-04 01:34:58 +02:00
if ( opt_drop )
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
fprintf ( sql_file , " /*!50003 DROP %s IF EXISTS %s */; \n " ,
2005-09-13 19:32:37 +02:00
routine_type [ i ] , routine_name ) ;
2006-03-02 13:18:49 +01:00
2007-08-03 09:50:17 +02:00
if ( mysql_num_fields ( routine_res ) > = 6 )
2006-03-02 13:18:49 +01:00
{
2007-08-03 09:50:17 +02:00
if ( switch_db_collation ( sql_file , db_name_buff , " ; " ,
db_cl_name , row [ 5 ] , & db_cl_altered ) )
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
{
2007-08-03 09:50:17 +02:00
DBUG_RETURN ( 1 ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
}
2006-03-02 13:18:49 +01:00
2007-08-03 09:50:17 +02:00
switch_cs_variables ( sql_file , " ; " ,
row [ 3 ] , /* character_set_client */
row [ 3 ] , /* character_set_results */
row [ 4 ] ) ; /* collation_connection */
2006-03-02 13:18:49 +01:00
}
2007-08-03 09:50:17 +02:00
else
2007-06-29 14:52:05 +02:00
{
2007-08-03 09:50:17 +02:00
/*
mysqldump is being run against the server , that does not
provide character set information in SHOW CREATE
statements .
NOTE : the dump may be incorrect , since character set
information is required in order to restore stored
procedure / function properly .
*/
fprintf ( sql_file ,
" -- \n "
" -- WARNING: old server version. "
" The following dump may be incomplete. \n "
" -- \n " ) ;
2007-06-29 14:52:05 +02:00
}
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
switch_sql_mode ( sql_file , " ; " , row [ 1 ] ) ;
2005-10-06 16:54:43 +02:00
fprintf ( sql_file ,
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
" DELIMITER ;; \n "
2012-11-19 17:11:35 +01:00
" %s ;; \n "
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
" DELIMITER ; \n " ,
2012-11-19 17:11:35 +01:00
( const char * ) row [ 2 ] ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
restore_sql_mode ( sql_file , " ; " ) ;
2007-08-06 20:42:13 +02:00
if ( mysql_num_fields ( routine_res ) > = 6 )
2007-06-29 14:52:05 +02:00
{
2007-08-03 09:50:17 +02:00
restore_cs_variables ( sql_file , " ; " ) ;
if ( db_cl_altered )
{
if ( restore_db_collation ( sql_file , db_name_buff , " ; " , db_cl_name ) )
DBUG_RETURN ( 1 ) ;
}
2007-06-29 14:52:05 +02:00
}
2006-03-02 13:18:49 +01:00
2005-09-04 01:34:58 +02:00
}
} /* end of routine printing */
2007-03-23 23:15:11 +01:00
mysql_free_result ( routine_res ) ;
2005-09-04 01:34:58 +02:00
} /* end of list of routines */
}
mysql_free_result ( routine_list_res ) ;
} /* end of for i (0 .. 1) */
2012-01-10 09:03:45 +01:00
if ( opt_xml )
{
fputs ( " \t </routines> \n " , sql_file ) ;
check_io ( sql_file ) ;
}
2007-07-12 10:49:39 +02:00
if ( switch_character_set_results ( mysql , default_charset ) )
DBUG_RETURN ( 1 ) ;
2005-09-13 19:32:37 +02:00
if ( lock_tables )
2009-11-24 14:54:59 +01:00
( void ) mysql_query_with_error_report ( mysql , 0 , " UNLOCK TABLES " ) ;
2005-09-04 01:34:58 +02:00
DBUG_RETURN ( 0 ) ;
}
2005-03-14 23:47:35 +01:00
2013-05-17 15:24:36 +02:00
/* general_log or slow_log tables under mysql database */
static inline my_bool general_log_or_slow_log_tables ( const char * db ,
const char * table )
{
return ( ! my_strcasecmp ( charset_info , db , " mysql " ) ) & &
( ! my_strcasecmp ( charset_info , table , " general_log " ) | |
! my_strcasecmp ( charset_info , table , " slow_log " ) ) ;
}
2000-07-31 21:29:14 +02:00
/*
2005-09-27 07:43:09 +02:00
get_table_structure - - retrievs database structure , prints out corresponding
CREATE statement and fills out insert_pat if the table is the type we will
be dumping .
ARGS
table - table name
db - db name
2006-05-25 10:00:28 +02:00
table_type - table type , e . g . " MyISAM " or " InnoDB " , but also " VIEW "
2005-09-27 07:43:09 +02:00
ignore_flag - what we must particularly ignore - see IGNORE_ defines above
2003-09-03 17:48:10 +02:00
RETURN
number of fields in table , 0 if error
2000-07-31 21:29:14 +02:00
*/
2003-09-03 17:48:10 +02:00
2005-09-27 07:43:09 +02:00
static uint get_table_structure ( char * table , char * db , char * table_type ,
char * ignore_flag )
2000-07-31 21:29:14 +02:00
{
2005-10-11 23:58:22 +02:00
my_bool init = 0 , delayed , write_data , complete_insert ;
2005-10-25 19:04:31 +02:00
my_ulonglong num_fields ;
2006-05-25 10:00:28 +02:00
char * result_table , * opt_quoted_table ;
2005-05-07 18:49:39 +02:00
const char * insert_option ;
2003-09-03 17:48:10 +02:00
char name_buff [ NAME_LEN + 3 ] , table_buff [ NAME_LEN * 2 + 3 ] ;
2006-06-21 14:14:37 +02:00
char table_buff2 [ NAME_LEN * 2 + 3 ] , query_buff [ QUERY_LENGTH ] ;
2011-01-14 15:20:34 +01:00
const char * show_fields_stmt = " SELECT `COLUMN_NAME` AS `Field`, "
" `COLUMN_TYPE` AS `Type`, "
" `IS_NULLABLE` AS `Null`, "
" `COLUMN_KEY` AS `Key`, "
" `COLUMN_DEFAULT` AS `Default`, "
" `EXTRA` AS `Extra`, "
" `COLUMN_COMMENT` AS `Comment` "
" FROM `INFORMATION_SCHEMA`.`COLUMNS` WHERE "
" TABLE_SCHEMA = '%s' AND TABLE_NAME = '%s' " ;
2006-08-02 03:18:47 +02:00
FILE * sql_file = md_result_file ;
2005-05-18 18:12:37 +02:00
int len ;
Bug #11754178 45740: MYSQLDUMP DOESN'T DUMP GENERAL_LOG AND SLOW_QUERY
CAUSES RESTORE PROBLEM
Problem Statement:
------------------
mysqldump is not having the dump stmts for general_log and slow_log
tables. That is because of the fix for Bug#26121. Hence, after
dropping the mysql database, and applying the dump by enabling the
logging, "'general_log' table not found" errors are logged into the
server log file.
Analysis:
---------
As part of the fix for Bug#26121, we skipped the dumping of tables
for general_log and slow_log, because the data dump of those tables
are taking LOCKS, which is not allowed for log tables.
Fix:
----
We came up with an approach that instead of taking both meta data
and data dump information for those tables, take only the meta data
dump which doesn't need LOCKS.
As part of fixing the issue we came up with below algorithm.
Design before fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
Design with the fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Explicitly call the 'show create table' for general_log and
slow_log
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
While taking the meta data dump for general_log and slow_log the
"CREATE TABLE" is replaced with "CREATE TABLE IF NOT EXISTS".
This is because we skipped "DROP TABLE" for those tables,
"DROP TABLE" fails for these tables if logging is enabled.
Customer is applying the dump by enabling logging so, if the dump
has "DROP TABLE" it will fail. Hence, removed the "DROP TABLE"
stmts for those tables.
After the fix we could observe "Table 'mysql.general_log'
doesn't exist" errors initially that is because in the customer
scenario they are dropping the mysql database by enabling the
logging, Hence, those errors are expected. Once we apply the
dump which is taken before the "drop database mysql", the errors
will not be there.
2012-05-04 15:03:34 +02:00
my_bool is_log_table ;
2005-10-25 19:04:31 +02:00
MYSQL_RES * result ;
MYSQL_ROW row ;
2005-04-04 23:32:48 +02:00
DBUG_ENTER ( " get_table_structure " ) ;
2005-10-11 23:58:22 +02:00
DBUG_PRINT ( " enter " , ( " db: %s table: %s " , db , table ) ) ;
2000-07-31 21:29:14 +02:00
2005-09-27 07:43:09 +02:00
* ignore_flag = check_if_ignore_table ( table , table_type ) ;
2005-10-11 23:58:22 +02:00
delayed = opt_delayed ;
if ( delayed & & ( * ignore_flag & IGNORE_INSERT_DELAYED ) )
{
delayed = 0 ;
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Warning: Unable to use delayed inserts for table '%s' "
" because it's of type %s \n " , table , table_type ) ;
2005-10-11 23:58:22 +02:00
}
2005-09-27 07:43:09 +02:00
2005-10-11 23:58:22 +02:00
complete_insert = 0 ;
if ( ( write_data = ! ( * ignore_flag & IGNORE_DATA ) ) )
2005-05-08 21:02:46 +02:00
{
2005-10-11 23:58:22 +02:00
complete_insert = opt_complete_insert ;
2005-09-27 07:43:09 +02:00
if ( ! insert_pat_inited )
2006-11-20 21:42:06 +01:00
{
insert_pat_inited = 1 ;
2007-03-22 18:35:29 +01:00
init_dynamic_string_checked ( & insert_pat , " " , 1024 , 1024 ) ;
2006-11-20 21:42:06 +01:00
}
2005-09-27 07:43:09 +02:00
else
2007-03-22 18:35:29 +01:00
dynstr_set_checked ( & insert_pat , " " ) ;
2005-05-08 21:02:46 +02:00
}
2005-10-11 23:58:22 +02:00
insert_option = ( ( delayed & & opt_ignore ) ? " DELAYED IGNORE " :
delayed ? " DELAYED " : opt_ignore ? " IGNORE " : " " ) ;
2000-07-31 21:29:14 +02:00
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Retrieving table structure for table %s... \n " , table ) ;
2000-07-31 21:29:14 +02:00
2005-05-18 18:12:37 +02:00
len = my_snprintf ( query_buff , sizeof ( query_buff ) ,
2013-07-18 08:10:08 +02:00
" SET SQL_QUOTE_SHOW_CREATE=%d " ,
2005-05-18 18:12:37 +02:00
( opt_quoted | | opt_keywords ) ) ;
2004-04-27 14:33:40 +02:00
if ( ! create_options )
2005-09-27 07:43:09 +02:00
strmov ( query_buff + len ,
" /*!40102 ,SQL_MODE=concat(@@sql_mode, _utf8 ',NO_KEY_OPTIONS,NO_TABLE_OPTIONS,NO_FIELD_OPTIONS') */ " ) ;
2004-04-27 14:33:40 +02:00
2003-09-03 17:48:10 +02:00
result_table = quote_name ( table , table_buff , 1 ) ;
opt_quoted_table = quote_name ( table , table_buff2 , 0 ) ;
2004-11-16 23:45:24 +01:00
if ( opt_order_by_primary )
2006-08-02 03:18:47 +02:00
order_by = primary_key_fields ( result_table ) ;
2004-11-16 23:45:24 +01:00
2006-07-24 12:00:44 +02:00
if ( ! opt_xml & & ! mysql_query_with_error_report ( mysql , 0 , query_buff ) )
2000-07-31 21:29:14 +02:00
{
2000-10-29 13:49:42 +01:00
/* using SHOW CREATE statement */
2006-07-24 11:56:01 +02:00
if ( ! opt_no_create_info )
2000-07-31 21:29:14 +02:00
{
2000-10-29 13:49:42 +01:00
/* Make an sql-file, if path was given iow. option -T was given */
char buff [ 20 + FN_REFLEN ] ;
2004-11-12 04:01:46 +01:00
MYSQL_FIELD * field ;
2000-10-29 13:49:42 +01:00
2005-03-16 21:56:56 +01:00
my_snprintf ( buff , sizeof ( buff ) , " show create table %s " , result_table ) ;
2007-07-25 17:46:50 +02:00
if ( switch_character_set_results ( mysql , " binary " ) | |
mysql_query_with_error_report ( mysql , & result , buff ) | |
switch_character_set_results ( mysql , default_charset ) )
2000-10-29 13:49:42 +01:00
DBUG_RETURN ( 0 ) ;
2000-07-31 21:29:14 +02:00
2000-10-29 13:49:42 +01:00
if ( path )
{
2009-05-27 15:16:21 +02:00
if ( ! ( sql_file = open_sql_file_for_table ( table , O_WRONLY ) ) )
2006-05-25 10:00:28 +02:00
DBUG_RETURN ( 0 ) ;
2007-03-22 18:35:29 +01:00
2002-01-02 20:29:41 +01:00
write_header ( sql_file , db ) ;
2000-10-29 13:49:42 +01:00
}
2012-01-10 09:03:45 +01:00
2006-05-25 10:00:28 +02:00
if ( strcmp ( table_type , " VIEW " ) = = 0 ) /* view */
2012-01-10 09:03:45 +01:00
print_comment ( sql_file , 0 ,
" \n -- \n -- Temporary table structure for view %s \n -- \n \n " ,
result_table ) ;
2006-05-25 10:00:28 +02:00
else
2012-01-10 09:03:45 +01:00
print_comment ( sql_file , 0 ,
" \n -- \n -- Table structure for table %s \n -- \n \n " ,
result_table ) ;
2000-10-29 13:49:42 +01:00
if ( opt_drop )
2004-06-24 22:14:12 +02:00
{
2006-05-25 10:00:28 +02:00
/*
Even if the " table " is a view , we do a DROP TABLE here . The
view - specific code below fills in the DROP VIEW .
Bug #11754178 45740: MYSQLDUMP DOESN'T DUMP GENERAL_LOG AND SLOW_QUERY
CAUSES RESTORE PROBLEM
Problem Statement:
------------------
mysqldump is not having the dump stmts for general_log and slow_log
tables. That is because of the fix for Bug#26121. Hence, after
dropping the mysql database, and applying the dump by enabling the
logging, "'general_log' table not found" errors are logged into the
server log file.
Analysis:
---------
As part of the fix for Bug#26121, we skipped the dumping of tables
for general_log and slow_log, because the data dump of those tables
are taking LOCKS, which is not allowed for log tables.
Fix:
----
We came up with an approach that instead of taking both meta data
and data dump information for those tables, take only the meta data
dump which doesn't need LOCKS.
As part of fixing the issue we came up with below algorithm.
Design before fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
Design with the fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Explicitly call the 'show create table' for general_log and
slow_log
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
While taking the meta data dump for general_log and slow_log the
"CREATE TABLE" is replaced with "CREATE TABLE IF NOT EXISTS".
This is because we skipped "DROP TABLE" for those tables,
"DROP TABLE" fails for these tables if logging is enabled.
Customer is applying the dump by enabling logging so, if the dump
has "DROP TABLE" it will fail. Hence, removed the "DROP TABLE"
stmts for those tables.
After the fix we could observe "Table 'mysql.general_log'
doesn't exist" errors initially that is because in the customer
scenario they are dropping the mysql database by enabling the
logging, Hence, those errors are expected. Once we apply the
dump which is taken before the "drop database mysql", the errors
will not be there.
2012-05-04 15:03:34 +02:00
We will skip the DROP TABLE for general_log and slow_log , since
those stmts will fail , in case we apply dump by enabling logging .
2006-05-25 10:00:28 +02:00
*/
Bug #11754178 45740: MYSQLDUMP DOESN'T DUMP GENERAL_LOG AND SLOW_QUERY
CAUSES RESTORE PROBLEM
Problem Statement:
------------------
mysqldump is not having the dump stmts for general_log and slow_log
tables. That is because of the fix for Bug#26121. Hence, after
dropping the mysql database, and applying the dump by enabling the
logging, "'general_log' table not found" errors are logged into the
server log file.
Analysis:
---------
As part of the fix for Bug#26121, we skipped the dumping of tables
for general_log and slow_log, because the data dump of those tables
are taking LOCKS, which is not allowed for log tables.
Fix:
----
We came up with an approach that instead of taking both meta data
and data dump information for those tables, take only the meta data
dump which doesn't need LOCKS.
As part of fixing the issue we came up with below algorithm.
Design before fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
Design with the fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Explicitly call the 'show create table' for general_log and
slow_log
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
While taking the meta data dump for general_log and slow_log the
"CREATE TABLE" is replaced with "CREATE TABLE IF NOT EXISTS".
This is because we skipped "DROP TABLE" for those tables,
"DROP TABLE" fails for these tables if logging is enabled.
Customer is applying the dump by enabling logging so, if the dump
has "DROP TABLE" it will fail. Hence, removed the "DROP TABLE"
stmts for those tables.
After the fix we could observe "Table 'mysql.general_log'
doesn't exist" errors initially that is because in the customer
scenario they are dropping the mysql database by enabling the
logging, Hence, those errors are expected. Once we apply the
dump which is taken before the "drop database mysql", the errors
will not be there.
2012-05-04 15:03:34 +02:00
if ( ! general_log_or_slow_log_tables ( db , table ) )
fprintf ( sql_file , " DROP TABLE IF EXISTS %s; \n " ,
opt_quoted_table ) ;
2006-05-25 10:00:28 +02:00
check_io ( sql_file ) ;
2004-06-24 22:14:12 +02:00
}
2000-10-29 13:49:42 +01:00
2005-10-25 19:04:31 +02:00
field = mysql_fetch_field_direct ( result , 0 ) ;
2004-11-12 04:01:46 +01:00
if ( strcmp ( field - > name , " View " ) = = 0 )
{
2006-08-02 03:18:47 +02:00
char * scv_buff = NULL ;
2012-07-04 16:48:58 +02:00
my_ulonglong n_cols ;
2006-05-30 14:49:05 +02:00
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- It's a view, create dummy table for view \n " ) ;
2005-06-16 15:58:17 +02:00
2006-05-30 14:49:05 +02:00
/* save "show create" statement for later */
if ( ( row = mysql_fetch_row ( result ) ) & & ( scv_buff = row [ 1 ] ) )
scv_buff = my_strdup ( scv_buff , MYF ( 0 ) ) ;
2005-10-25 19:04:31 +02:00
mysql_free_result ( result ) ;
/*
2006-05-25 10:00:28 +02:00
Create a table with the same name as the view and with columns of
2005-10-25 19:04:31 +02:00
the same name in order to satisfy views that depend on this view .
The table will be removed when the actual view is created .
2005-06-16 15:58:17 +02:00
2012-07-04 16:48:58 +02:00
The properties of each column , are not preserved in this temporary
table , because they are not necessary .
2005-06-16 15:58:17 +02:00
2005-10-25 19:04:31 +02:00
This will not be necessary once we can determine dependencies
between views and can simply dump them in the appropriate order .
*/
2005-06-16 15:58:17 +02:00
my_snprintf ( query_buff , sizeof ( query_buff ) ,
2005-10-25 19:04:31 +02:00
" SHOW FIELDS FROM %s " , result_table ) ;
2007-07-27 16:20:17 +02:00
if ( switch_character_set_results ( mysql , " binary " ) | |
mysql_query_with_error_report ( mysql , & result , query_buff ) | |
switch_character_set_results ( mysql , default_charset ) )
2005-06-16 15:58:17 +02:00
{
2006-05-30 14:49:05 +02:00
/*
View references invalid or privileged table / col / fun ( err 1356 ) ,
so we cannot create a stand - in table . Be defensive and dump
a comment with the view ' s ' show create ' statement . ( Bug # 17371 )
*/
2006-07-24 12:00:44 +02:00
if ( mysql_errno ( mysql ) = = ER_VIEW_INVALID )
2006-05-30 14:49:05 +02:00
fprintf ( sql_file , " \n -- failed on view %s: %s \n \n " , result_table , scv_buff ? scv_buff : " " ) ;
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( scv_buff ) ;
2006-05-30 14:49:05 +02:00
2006-08-02 03:18:47 +02:00
DBUG_RETURN ( 0 ) ;
2005-06-16 15:58:17 +02:00
}
2006-05-30 14:49:05 +02:00
else
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( scv_buff ) ;
2005-06-16 15:58:17 +02:00
2012-07-04 16:48:58 +02:00
n_cols = mysql_num_rows ( result ) ;
if ( 0 ! = n_cols )
2005-06-16 15:58:17 +02:00
{
2012-07-04 16:48:58 +02:00
/*
The actual formula is based on the column names and how the . FRM
files are stored and is too volatile to be repeated here .
Thus we simply warn the user if the columns exceed a limit we
know works most of the time .
*/
if ( n_cols > = 1000 )
fprintf ( stderr ,
" -- Warning: Creating a stand-in table for view %s may "
" fail when replaying the dump file produced because "
" of the number of columns exceeding 1000. Exercise "
" caution when replaying the produced dump file. \n " ,
table ) ;
2007-07-27 16:20:17 +02:00
if ( opt_drop )
2005-10-25 19:04:31 +02:00
{
2006-05-25 10:00:28 +02:00
/*
2007-07-27 16:20:17 +02:00
We have already dropped any table of the same name above , so
here we just drop the view .
*/
2006-05-25 10:00:28 +02:00
2007-07-27 16:20:17 +02:00
fprintf ( sql_file , " /*!50001 DROP VIEW IF EXISTS %s*/; \n " ,
opt_quoted_table ) ;
check_io ( sql_file ) ;
}
2005-06-16 15:58:17 +02:00
2007-07-27 16:20:17 +02:00
fprintf ( sql_file ,
" SET @saved_cs_client = @@character_set_client; \n "
" SET character_set_client = utf8; \n "
" /*!50001 CREATE TABLE %s ( \n " ,
result_table ) ;
2005-10-25 19:04:31 +02:00
2007-07-27 16:20:17 +02:00
/*
Get first row , following loop will prepend comma - keeps from
having to know if the row being printed is last to determine if
there should be a _trailing_ comma .
*/
2005-10-25 19:04:31 +02:00
2007-07-27 16:20:17 +02:00
row = mysql_fetch_row ( result ) ;
2012-07-04 16:48:58 +02:00
/*
The actual column type doesn ' t matter anyway , since the table will
be dropped at run time .
We do tinyint to avoid hitting the row size limit .
*/
fprintf ( sql_file , " %s tinyint NOT NULL " ,
quote_name ( row [ 0 ] , name_buff , 0 ) ) ;
2007-07-27 16:20:17 +02:00
while ( ( row = mysql_fetch_row ( result ) ) )
{
/* col name, col type */
2012-07-04 16:48:58 +02:00
fprintf ( sql_file , " , \n %s tinyint NOT NULL " ,
quote_name ( row [ 0 ] , name_buff , 0 ) ) ;
2005-10-25 19:04:31 +02:00
}
2008-09-11 08:14:19 +02:00
/*
Stand - in tables are always MyISAM tables as the default
engine might have a column - limit that ' s lower than the
number of columns in the view , and MyISAM support is
guaranteed to be in the server anyway .
*/
2007-07-27 16:20:17 +02:00
fprintf ( sql_file ,
2008-09-11 08:14:19 +02:00
" \n ) ENGINE=MyISAM */; \n "
2007-07-27 16:20:17 +02:00
" SET character_set_client = @saved_cs_client; \n " ) ;
check_io ( sql_file ) ;
2005-06-16 15:58:17 +02:00
}
2007-07-27 16:20:17 +02:00
2005-10-25 19:04:31 +02:00
mysql_free_result ( result ) ;
2006-05-31 13:36:28 +02:00
if ( path )
my_fclose ( sql_file , MYF ( MY_WME ) ) ;
2006-05-25 10:00:28 +02:00
seen_views = 1 ;
2004-11-12 04:01:46 +01:00
DBUG_RETURN ( 0 ) ;
}
2005-10-25 19:04:31 +02:00
row = mysql_fetch_row ( result ) ;
2007-07-25 17:46:50 +02:00
Bug #11754178 45740: MYSQLDUMP DOESN'T DUMP GENERAL_LOG AND SLOW_QUERY
CAUSES RESTORE PROBLEM
Problem Statement:
------------------
mysqldump is not having the dump stmts for general_log and slow_log
tables. That is because of the fix for Bug#26121. Hence, after
dropping the mysql database, and applying the dump by enabling the
logging, "'general_log' table not found" errors are logged into the
server log file.
Analysis:
---------
As part of the fix for Bug#26121, we skipped the dumping of tables
for general_log and slow_log, because the data dump of those tables
are taking LOCKS, which is not allowed for log tables.
Fix:
----
We came up with an approach that instead of taking both meta data
and data dump information for those tables, take only the meta data
dump which doesn't need LOCKS.
As part of fixing the issue we came up with below algorithm.
Design before fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
Design with the fix:
1) mysql database is having tables like db, event,... general_log,
... slow_log...
2) Skip general_log and slow_log while preparing the tables list
3) Explicitly call the 'show create table' for general_log and
slow_log
3) Take the TL_READ lock on tables which are present in the table
list and do 'show create table'.
4) Release the lock.
While taking the meta data dump for general_log and slow_log the
"CREATE TABLE" is replaced with "CREATE TABLE IF NOT EXISTS".
This is because we skipped "DROP TABLE" for those tables,
"DROP TABLE" fails for these tables if logging is enabled.
Customer is applying the dump by enabling logging so, if the dump
has "DROP TABLE" it will fail. Hence, removed the "DROP TABLE"
stmts for those tables.
After the fix we could observe "Table 'mysql.general_log'
doesn't exist" errors initially that is because in the customer
scenario they are dropping the mysql database by enabling the
logging, Hence, those errors are expected. Once we apply the
dump which is taken before the "drop database mysql", the errors
will not be there.
2012-05-04 15:03:34 +02:00
is_log_table = general_log_or_slow_log_tables ( db , table ) ;
if ( is_log_table )
row [ 1 ] + = 13 ; /* strlen("CREATE TABLE ")= 13 */
if ( opt_compatible_mode & 3 )
{
fprintf ( sql_file ,
is_log_table ? " CREATE TABLE IF NOT EXISTS %s; \n " : " %s; \n " ,
row [ 1 ] ) ;
}
else
{
fprintf ( sql_file ,
" /*!40101 SET @saved_cs_client = @@character_set_client */; \n "
" /*!40101 SET character_set_client = utf8 */; \n "
" %s%s; \n "
" /*!40101 SET character_set_client = @saved_cs_client */; \n " ,
is_log_table ? " CREATE TABLE IF NOT EXISTS " : " " ,
row [ 1 ] ) ;
}
2007-07-25 17:46:50 +02:00
2004-06-24 22:14:12 +02:00
check_io ( sql_file ) ;
2005-10-25 19:04:31 +02:00
mysql_free_result ( result ) ;
2000-10-29 13:49:42 +01:00
}
2005-05-08 21:02:46 +02:00
my_snprintf ( query_buff , sizeof ( query_buff ) , " show fields from %s " ,
2006-05-25 10:00:28 +02:00
result_table ) ;
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & result , query_buff ) )
2000-07-31 21:29:14 +02:00
{
2002-01-02 20:29:41 +01:00
if ( path )
2006-05-25 10:00:28 +02:00
my_fclose ( sql_file , MYF ( MY_WME ) ) ;
2000-10-29 13:49:42 +01:00
DBUG_RETURN ( 0 ) ;
2000-07-31 21:29:14 +02:00
}
2000-10-29 13:49:42 +01:00
2005-09-27 07:43:09 +02:00
/*
2005-10-11 23:58:22 +02:00
If write_data is true , then we build up insert statements for
the table ' s data . Note : in subsequent lines of code , this test
will have to be performed each time we are appending to
insert_pat .
2005-09-27 07:43:09 +02:00
*/
2005-10-11 23:58:22 +02:00
if ( write_data )
2000-07-31 21:29:14 +02:00
{
2005-11-24 18:56:40 +01:00
if ( opt_replace_into )
2007-03-23 01:08:05 +01:00
dynstr_append_checked ( & insert_pat , " REPLACE " ) ;
2005-11-24 18:56:40 +01:00
else
2007-03-23 01:08:05 +01:00
dynstr_append_checked ( & insert_pat , " INSERT " ) ;
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , insert_option ) ;
dynstr_append_checked ( & insert_pat , " INTO " ) ;
dynstr_append_checked ( & insert_pat , opt_quoted_table ) ;
2005-10-11 23:58:22 +02:00
if ( complete_insert )
2005-09-27 07:43:09 +02:00
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " ( " ) ;
2005-09-27 07:43:09 +02:00
}
else
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " VALUES " ) ;
2005-09-27 07:43:09 +02:00
if ( ! extended_insert )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " ( " ) ;
2005-09-27 07:43:09 +02:00
}
2000-10-29 13:49:42 +01:00
}
2005-10-25 19:04:31 +02:00
while ( ( row = mysql_fetch_row ( result ) ) )
2000-10-29 13:49:42 +01:00
{
2005-10-11 23:58:22 +02:00
if ( complete_insert )
2000-07-31 21:29:14 +02:00
{
2005-10-11 23:58:22 +02:00
if ( init )
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " , " ) ;
2005-10-11 23:58:22 +02:00
}
init = 1 ;
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat ,
2005-05-08 21:02:46 +02:00
quote_name ( row [ SHOW_FIELDNAME ] , name_buff , 0 ) ) ;
2005-10-11 23:58:22 +02:00
}
2000-07-31 21:29:14 +02:00
}
2005-10-25 19:04:31 +02:00
num_fields = mysql_num_rows ( result ) ;
mysql_free_result ( result ) ;
2000-07-31 21:29:14 +02:00
}
2000-10-29 13:49:42 +01:00
else
2000-07-31 21:29:14 +02:00
{
2006-07-24 11:51:52 +02:00
verbose_msg ( " %s: Warning: Can't set SQL_QUOTE_SHOW_CREATE option (%s) \n " ,
2006-07-24 12:00:44 +02:00
my_progname , mysql_error ( mysql ) ) ;
2000-10-29 13:49:42 +01:00
2011-01-14 15:20:34 +01:00
my_snprintf ( query_buff , sizeof ( query_buff ) , show_fields_stmt , db , table ) ;
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & result , query_buff ) )
2000-07-31 21:29:14 +02:00
DBUG_RETURN ( 0 ) ;
2000-10-29 13:49:42 +01:00
/* Make an sql-file, if path was given iow. option -T was given */
2006-07-24 11:56:01 +02:00
if ( ! opt_no_create_info )
2000-07-31 21:29:14 +02:00
{
2000-10-29 13:49:42 +01:00
if ( path )
2000-07-31 21:29:14 +02:00
{
2009-05-27 15:16:21 +02:00
if ( ! ( sql_file = open_sql_file_for_table ( table , O_WRONLY ) ) )
2006-05-25 10:00:28 +02:00
DBUG_RETURN ( 0 ) ;
2002-01-02 20:29:41 +01:00
write_header ( sql_file , db ) ;
2000-07-31 21:29:14 +02:00
}
2012-01-10 09:03:45 +01:00
print_comment ( sql_file , 0 ,
" \n -- \n -- Table structure for table %s \n -- \n \n " ,
result_table ) ;
2000-10-29 13:49:42 +01:00
if ( opt_drop )
2005-07-18 14:33:18 +02:00
fprintf ( sql_file , " DROP TABLE IF EXISTS %s; \n " , result_table ) ;
2003-10-30 11:58:30 +01:00
if ( ! opt_xml )
2006-05-25 10:00:28 +02:00
fprintf ( sql_file , " CREATE TABLE %s ( \n " , result_table ) ;
2003-10-30 11:58:30 +01:00
else
2006-10-19 00:43:51 +02:00
print_xml_tag ( sql_file , " \t " , " \n " , " table_structure " , " name= " , table ,
NullS ) ;
2004-06-24 22:14:12 +02:00
check_io ( sql_file ) ;
2000-10-29 13:49:42 +01:00
}
2005-05-08 21:02:46 +02:00
2005-10-11 23:58:22 +02:00
if ( write_data )
2005-05-08 21:02:46 +02:00
{
2005-11-24 18:56:40 +01:00
if ( opt_replace_into )
2007-03-23 01:08:05 +01:00
dynstr_append_checked ( & insert_pat , " REPLACE " ) ;
2005-11-24 18:56:40 +01:00
else
2007-03-23 01:08:05 +01:00
dynstr_append_checked ( & insert_pat , " INSERT " ) ;
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , insert_option ) ;
dynstr_append_checked ( & insert_pat , " INTO " ) ;
dynstr_append_checked ( & insert_pat , result_table ) ;
2007-03-26 09:56:41 +02:00
if ( complete_insert )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " ( " ) ;
2005-09-27 07:43:09 +02:00
else
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " VALUES " ) ;
2005-09-27 07:43:09 +02:00
if ( ! extended_insert )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " ( " ) ;
2005-09-27 07:43:09 +02:00
}
2000-07-31 21:29:14 +02:00
}
2000-10-29 13:49:42 +01:00
2005-10-25 19:04:31 +02:00
while ( ( row = mysql_fetch_row ( result ) ) )
2000-07-31 21:29:14 +02:00
{
2005-10-25 19:04:31 +02:00
ulong * lengths = mysql_fetch_lengths ( result ) ;
2000-10-29 13:49:42 +01:00
if ( init )
2000-07-31 21:29:14 +02:00
{
2006-07-24 11:56:01 +02:00
if ( ! opt_xml & & ! opt_no_create_info )
2006-05-25 10:00:28 +02:00
{
fputs ( " , \n " , sql_file ) ;
check_io ( sql_file ) ;
}
2005-10-11 23:58:22 +02:00
if ( complete_insert )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " , " ) ;
2000-10-29 13:49:42 +01:00
}
init = 1 ;
2007-03-26 09:56:41 +02:00
if ( complete_insert )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat ,
2005-05-08 21:02:46 +02:00
quote_name ( row [ SHOW_FIELDNAME ] , name_buff , 0 ) ) ;
2006-07-24 11:56:01 +02:00
if ( ! opt_no_create_info )
2000-10-29 13:49:42 +01:00
{
2006-05-25 10:00:28 +02:00
if ( opt_xml )
{
2012-01-10 09:03:45 +01:00
print_xml_row ( sql_file , " field " , result , & row , NullS ) ;
2006-05-25 10:00:28 +02:00
continue ;
}
2005-05-08 21:02:46 +02:00
2000-10-29 13:49:42 +01:00
if ( opt_keywords )
2006-05-25 10:00:28 +02:00
fprintf ( sql_file , " %s.%s %s " , result_table ,
quote_name ( row [ SHOW_FIELDNAME ] , name_buff , 0 ) ,
row [ SHOW_TYPE ] ) ;
2000-10-29 13:49:42 +01:00
else
2006-05-25 10:00:28 +02:00
fprintf ( sql_file , " %s %s " , quote_name ( row [ SHOW_FIELDNAME ] ,
name_buff , 0 ) ,
row [ SHOW_TYPE ] ) ;
2000-10-29 13:49:42 +01:00
if ( row [ SHOW_DEFAULT ] )
{
2006-05-25 10:00:28 +02:00
fputs ( " DEFAULT " , sql_file ) ;
unescape ( sql_file , row [ SHOW_DEFAULT ] , lengths [ SHOW_DEFAULT ] ) ;
2000-10-29 13:49:42 +01:00
}
if ( ! row [ SHOW_NULL ] [ 0 ] )
2006-05-25 10:00:28 +02:00
fputs ( " NOT NULL " , sql_file ) ;
2000-10-29 13:49:42 +01:00
if ( row [ SHOW_EXTRA ] [ 0 ] )
2006-05-25 10:00:28 +02:00
fprintf ( sql_file , " %s " , row [ SHOW_EXTRA ] ) ;
check_io ( sql_file ) ;
2000-07-31 21:29:14 +02:00
}
}
2005-10-25 19:04:31 +02:00
num_fields = mysql_num_rows ( result ) ;
mysql_free_result ( result ) ;
2006-07-24 11:56:01 +02:00
if ( ! opt_no_create_info )
2000-07-31 21:29:14 +02:00
{
2000-10-29 13:49:42 +01:00
/* Make an sql-file, if path was given iow. option -T was given */
char buff [ 20 + FN_REFLEN ] ;
uint keynr , primary_key ;
2005-03-16 21:56:56 +01:00
my_snprintf ( buff , sizeof ( buff ) , " show keys from %s " , result_table ) ;
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & result , buff ) )
2000-07-31 21:29:14 +02:00
{
2006-07-24 12:00:44 +02:00
if ( mysql_errno ( mysql ) = = ER_WRONG_OBJECT )
2004-11-12 04:01:46 +01:00
{
/* it is VIEW */
fputs ( " \t \t <options Comment= \" view \" /> \n " , sql_file ) ;
goto continue_xml ;
}
fprintf ( stderr , " %s: Can't get keys for table %s (%s) \n " ,
2006-07-24 12:00:44 +02:00
my_progname , result_table , mysql_error ( mysql ) ) ;
2002-01-02 20:29:41 +01:00
if ( path )
2006-05-25 10:00:28 +02:00
my_fclose ( sql_file , MYF ( MY_WME ) ) ;
2000-10-29 13:49:42 +01:00
DBUG_RETURN ( 0 ) ;
2000-07-31 21:29:14 +02:00
}
2000-10-29 13:49:42 +01:00
/* Find first which key is primary key */
keynr = 0 ;
primary_key = INT_MAX ;
2005-10-25 19:04:31 +02:00
while ( ( row = mysql_fetch_row ( result ) ) )
2000-07-31 21:29:14 +02:00
{
2000-10-29 13:49:42 +01:00
if ( atoi ( row [ 3 ] ) = = 1 )
{
2006-05-25 10:00:28 +02:00
keynr + + ;
2001-04-23 23:31:34 +02:00
# ifdef FORCE_PRIMARY_KEY
2006-05-25 10:00:28 +02:00
if ( atoi ( row [ 1 ] ) = = 0 & & primary_key = = INT_MAX )
primary_key = keynr ;
2001-04-23 23:31:34 +02:00
# endif
2006-05-25 10:00:28 +02:00
if ( ! strcmp ( row [ 2 ] , " PRIMARY " ) )
{
primary_key = keynr ;
break ;
}
2000-10-29 13:49:42 +01:00
}
2000-07-31 21:29:14 +02:00
}
2005-10-25 19:04:31 +02:00
mysql_data_seek ( result , 0 ) ;
2000-10-29 13:49:42 +01:00
keynr = 0 ;
2005-10-25 19:04:31 +02:00
while ( ( row = mysql_fetch_row ( result ) ) )
2000-10-29 13:49:42 +01:00
{
2006-05-25 10:00:28 +02:00
if ( opt_xml )
{
2012-01-10 09:03:45 +01:00
print_xml_row ( sql_file , " key " , result , & row , NullS ) ;
2006-05-25 10:00:28 +02:00
continue ;
}
2005-09-27 07:43:09 +02:00
2000-10-29 13:49:42 +01:00
if ( atoi ( row [ 3 ] ) = = 1 )
{
2006-05-25 10:00:28 +02:00
if ( keynr + + )
putc ( ' ) ' , sql_file ) ;
if ( atoi ( row [ 1 ] ) ) /* Test if duplicate key */
/* Duplicate allowed */
fprintf ( sql_file , " , \n KEY %s ( " , quote_name ( row [ 2 ] , name_buff , 0 ) ) ;
else if ( keynr = = primary_key )
fputs ( " , \n PRIMARY KEY ( " , sql_file ) ; /* First UNIQUE is primary */
else
fprintf ( sql_file , " , \n UNIQUE %s ( " , quote_name ( row [ 2 ] , name_buff ,
0 ) ) ;
2000-10-29 13:49:42 +01:00
}
else
2006-05-25 10:00:28 +02:00
putc ( ' , ' , sql_file ) ;
2003-09-03 17:48:10 +02:00
fputs ( quote_name ( row [ 4 ] , name_buff , 0 ) , sql_file ) ;
2000-10-29 13:49:42 +01:00
if ( row [ 7 ] )
2006-05-25 10:00:28 +02:00
fprintf ( sql_file , " (%s) " , row [ 7 ] ) ; /* Sub key */
check_io ( sql_file ) ;
2000-10-29 13:49:42 +01:00
}
2007-08-02 06:49:29 +02:00
mysql_free_result ( result ) ;
2003-10-30 11:58:30 +01:00
if ( ! opt_xml )
{
2006-05-25 10:00:28 +02:00
if ( keynr )
putc ( ' ) ' , sql_file ) ;
fputs ( " \n ) " , sql_file ) ;
check_io ( sql_file ) ;
2003-10-30 11:58:30 +01:00
}
2000-10-29 13:49:42 +01:00
/* Get MySQL specific create options */
if ( create_options )
2000-07-31 21:29:14 +02:00
{
2006-05-25 10:00:28 +02:00
char show_name_buff [ NAME_LEN * 2 + 2 + 24 ] ;
2005-03-16 21:56:56 +01:00
2006-05-25 10:00:28 +02:00
/* Check memory for quote_for_like() */
2005-03-16 21:56:56 +01:00
my_snprintf ( buff , sizeof ( buff ) , " show table status like %s " ,
2006-05-25 10:00:28 +02:00
quote_for_like ( table , show_name_buff ) ) ;
2005-03-16 21:56:56 +01:00
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & result , buff ) )
2000-10-29 13:49:42 +01:00
{
2006-07-24 12:00:44 +02:00
if ( mysql_errno ( mysql ) ! = ER_PARSE_ERROR )
2006-05-25 10:00:28 +02:00
{ /* If old MySQL version */
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Warning: Couldn't get status information for " \
2006-07-24 12:00:44 +02:00
" table %s (%s) \n " , result_table , mysql_error ( mysql ) ) ;
2006-05-25 10:00:28 +02:00
}
2000-10-29 13:49:42 +01:00
}
2005-10-25 19:04:31 +02:00
else if ( ! ( row = mysql_fetch_row ( result ) ) )
2000-10-29 13:49:42 +01:00
{
2006-05-25 10:00:28 +02:00
fprintf ( stderr ,
" Error: Couldn't read status information for table %s (%s) \n " ,
2006-07-24 12:00:44 +02:00
result_table , mysql_error ( mysql ) ) ;
2000-10-29 13:49:42 +01:00
}
else
{
2006-05-25 10:00:28 +02:00
if ( opt_xml )
2012-01-10 09:03:45 +01:00
print_xml_row ( sql_file , " options " , result , & row , NullS ) ;
2006-05-25 10:00:28 +02:00
else
{
fputs ( " /*! " , sql_file ) ;
print_value ( sql_file , result , row , " engine= " , " Engine " , 0 ) ;
print_value ( sql_file , result , row , " " , " Create_options " , 0 ) ;
print_value ( sql_file , result , row , " comment= " , " Comment " , 1 ) ;
fputs ( " */ " , sql_file ) ;
check_io ( sql_file ) ;
}
2000-10-29 13:49:42 +01:00
}
2006-05-25 10:00:28 +02:00
mysql_free_result ( result ) ; /* Is always safe to free */
2000-07-31 21:29:14 +02:00
}
2004-11-12 04:01:46 +01:00
continue_xml :
2003-10-30 11:58:30 +01:00
if ( ! opt_xml )
2006-05-25 10:00:28 +02:00
fputs ( " ; \n " , sql_file ) ;
2003-10-30 11:58:30 +01:00
else
2006-05-25 10:00:28 +02:00
fputs ( " \t </table_structure> \n " , sql_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( sql_file ) ;
2000-07-31 21:29:14 +02:00
}
}
2007-03-26 09:56:41 +02:00
if ( complete_insert )
2000-07-31 21:29:14 +02:00
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " ) VALUES " ) ;
2000-07-31 21:29:14 +02:00
if ( ! extended_insert )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & insert_pat , " ( " ) ;
2000-07-31 21:29:14 +02:00
}
2002-01-02 20:29:41 +01:00
if ( sql_file ! = md_result_file )
2004-02-13 22:21:46 +01:00
{
fputs ( " \n " , sql_file ) ;
write_footer ( sql_file ) ;
2002-01-02 20:29:41 +01:00
my_fclose ( sql_file , MYF ( MY_WME ) ) ;
2004-02-13 22:21:46 +01:00
}
2006-11-30 02:40:42 +01:00
DBUG_RETURN ( ( uint ) num_fields ) ;
2005-04-04 23:32:48 +02:00
} /* get_table_structure */
2000-07-31 21:29:14 +02:00
2009-05-27 15:16:21 +02:00
static void dump_trigger_old ( FILE * sql_file , MYSQL_RES * show_triggers_rs ,
2007-08-03 09:50:17 +02:00
MYSQL_ROW * show_trigger_row ,
const char * table_name )
{
char quoted_table_name_buf [ NAME_LEN * 2 + 3 ] ;
char * quoted_table_name = quote_name ( table_name , quoted_table_name_buf , 1 ) ;
char name_buff [ NAME_LEN * 4 + 3 ] ;
2012-01-10 09:03:45 +01:00
const char * xml_msg = " \n Warning! mysqldump being run against old server "
" that does not \n support 'SHOW CREATE TRIGGERS' "
" statement. Skipping.. \n " ;
2007-08-03 09:50:17 +02:00
DBUG_ENTER ( " dump_trigger_old " ) ;
2012-01-10 09:03:45 +01:00
if ( opt_xml )
{
print_xml_comment ( sql_file , strlen ( xml_msg ) , xml_msg ) ;
check_io ( sql_file ) ;
DBUG_VOID_RETURN ;
}
2007-08-03 09:50:17 +02:00
fprintf ( sql_file ,
" -- \n "
" -- WARNING: old server version. "
" The following dump may be incomplete. \n "
" -- \n " ) ;
if ( opt_compact )
fprintf ( sql_file , " /*!50003 SET @OLD_SQL_MODE=@@SQL_MODE*/; \n " ) ;
fprintf ( sql_file ,
" DELIMITER ;; \n "
" /*!50003 SET SESSION SQL_MODE= \" %s \" */;; \n "
" /*!50003 CREATE */ " ,
( * show_trigger_row ) [ 6 ] ) ;
if ( mysql_num_fields ( show_triggers_rs ) > 7 )
{
/*
mysqldump can be run against the server , that does not support
definer in triggers ( there is no DEFINER column in SHOW TRIGGERS
output ) . So , we should check if we have this column before
accessing it .
*/
2005-09-09 00:03:07 +02:00
2007-08-06 20:49:47 +02:00
size_t user_name_len ;
2007-08-03 09:50:17 +02:00
char user_name_str [ USERNAME_LENGTH + 1 ] ;
char quoted_user_name_str [ USERNAME_LENGTH * 2 + 3 ] ;
2007-08-06 20:49:47 +02:00
size_t host_name_len ;
2007-08-03 09:50:17 +02:00
char host_name_str [ HOSTNAME_LENGTH + 1 ] ;
char quoted_host_name_str [ HOSTNAME_LENGTH * 2 + 3 ] ;
2005-09-09 00:03:07 +02:00
2007-08-03 09:50:17 +02:00
parse_user ( ( * show_trigger_row ) [ 7 ] ,
strlen ( ( * show_trigger_row ) [ 7 ] ) ,
user_name_str , & user_name_len ,
host_name_str , & host_name_len ) ;
2005-09-09 00:03:07 +02:00
2007-08-03 09:50:17 +02:00
fprintf ( sql_file ,
" /*!50017 DEFINER=%s@%s */ " ,
quote_name ( user_name_str , quoted_user_name_str , FALSE ) ,
quote_name ( host_name_str , quoted_host_name_str , FALSE ) ) ;
}
fprintf ( sql_file ,
" /*!50003 TRIGGER %s %s %s ON %s FOR EACH ROW%s%s */;; \n "
" DELIMITER ; \n " ,
quote_name ( ( * show_trigger_row ) [ 0 ] , name_buff , 0 ) , /* Trigger */
( * show_trigger_row ) [ 4 ] , /* Timing */
( * show_trigger_row ) [ 1 ] , /* Event */
quoted_table_name ,
( strchr ( " \t \n \r " , * ( ( * show_trigger_row ) [ 3 ] ) ) ) ? " " : " " ,
( * show_trigger_row ) [ 3 ] /* Statement */ ) ;
if ( opt_compact )
fprintf ( sql_file , " /*!50003 SET SESSION SQL_MODE=@OLD_SQL_MODE */; \n " ) ;
DBUG_VOID_RETURN ;
}
2009-05-27 15:16:21 +02:00
static int dump_trigger ( FILE * sql_file , MYSQL_RES * show_create_trigger_rs ,
2007-08-03 09:50:17 +02:00
const char * db_name ,
const char * db_cl_name )
{
MYSQL_ROW row ;
2012-01-10 09:03:45 +01:00
char * query_str ;
2007-08-06 20:52:50 +02:00
int db_cl_altered = FALSE ;
2005-09-09 00:03:07 +02:00
2007-08-03 09:50:17 +02:00
DBUG_ENTER ( " dump_trigger " ) ;
while ( ( row = mysql_fetch_row ( show_create_trigger_rs ) ) )
{
2012-01-10 09:03:45 +01:00
if ( opt_xml )
{
print_xml_row ( sql_file , " trigger " , show_create_trigger_rs , & row ,
" SQL Original Statement " ) ;
check_io ( sql_file ) ;
continue ;
}
2007-08-03 09:50:17 +02:00
2012-01-10 11:40:48 +01:00
query_str = cover_definer_clause ( row [ 2 ] , strlen ( row [ 2 ] ) ,
C_STRING_WITH_LEN ( " 50017 " ) ,
C_STRING_WITH_LEN ( " 50003 " ) ,
C_STRING_WITH_LEN ( " TRIGGER " ) ) ;
2007-08-03 09:50:17 +02:00
if ( switch_db_collation ( sql_file , db_name , " ; " ,
db_cl_name , row [ 5 ] , & db_cl_altered ) )
DBUG_RETURN ( TRUE ) ;
switch_cs_variables ( sql_file , " ; " ,
row [ 3 ] , /* character_set_client */
row [ 3 ] , /* character_set_results */
row [ 4 ] ) ; /* collation_connection */
switch_sql_mode ( sql_file , " ; " , row [ 1 ] ) ;
fprintf ( sql_file ,
" DELIMITER ;; \n "
" /*!50003 %s */;; \n "
" DELIMITER ; \n " ,
( const char * ) ( query_str ! = NULL ? query_str : row [ 2 ] ) ) ;
restore_sql_mode ( sql_file , " ; " ) ;
restore_cs_variables ( sql_file , " ; " ) ;
if ( db_cl_altered )
{
if ( restore_db_collation ( sql_file , db_name , " ; " , db_cl_name ) )
DBUG_RETURN ( TRUE ) ;
}
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( query_str ) ;
2007-08-03 09:50:17 +02:00
}
DBUG_RETURN ( FALSE ) ;
}
/**
Dump the triggers for a given table .
This should be called after the tables have been dumped in case a trigger
depends on the existence of a table .
@ param [ in ] table_name
@ param [ in ] db_name
@ return Error status .
@ retval TRUE error has occurred .
@ retval FALSE operation succeed .
2005-09-09 00:03:07 +02:00
*/
2007-08-03 09:50:17 +02:00
static int dump_triggers_for_table ( char * table_name , char * db_name )
2005-09-09 00:03:07 +02:00
{
2007-08-03 09:50:17 +02:00
char name_buff [ NAME_LEN * 4 + 3 ] ;
2006-03-09 21:12:43 +01:00
char query_buff [ QUERY_LENGTH ] ;
2007-08-03 09:50:17 +02:00
uint old_opt_compatible_mode = opt_compatible_mode ;
MYSQL_RES * show_triggers_rs ;
2005-10-13 22:42:56 +02:00
MYSQL_ROW row ;
2009-05-27 15:16:21 +02:00
FILE * sql_file = md_result_file ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
char db_cl_name [ MY_CS_NAME_SIZE ] ;
2009-05-27 15:16:21 +02:00
int ret = TRUE ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2005-09-09 00:03:07 +02:00
DBUG_ENTER ( " dump_triggers_for_table " ) ;
2007-08-03 09:50:17 +02:00
DBUG_PRINT ( " enter " , ( " db: %s, table_name: %s " , db_name , table_name ) ) ;
2005-10-13 22:42:56 +02:00
2009-05-27 15:16:21 +02:00
if ( path & & ! ( sql_file = open_sql_file_for_table ( table_name ,
O_WRONLY | O_APPEND ) ) )
DBUG_RETURN ( 1 ) ;
2005-10-13 22:42:56 +02:00
/* Do not use ANSI_QUOTES on triggers in dump */
opt_compatible_mode & = ~ MASK_ANSI_QUOTES ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
/* Get database collation. */
2007-08-06 20:42:13 +02:00
if ( switch_character_set_results ( mysql , " binary " ) )
2009-05-27 15:16:21 +02:00
goto done ;
2007-08-06 20:42:13 +02:00
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
if ( fetch_db_collation ( db_name , db_cl_name , sizeof ( db_cl_name ) ) )
2009-05-27 15:16:21 +02:00
goto done ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-08-03 09:50:17 +02:00
/* Get list of triggers. */
my_snprintf ( query_buff , sizeof ( query_buff ) ,
" SHOW TRIGGERS LIKE %s " ,
quote_for_like ( table_name , name_buff ) ) ;
if ( mysql_query_with_error_report ( mysql , & show_triggers_rs , query_buff ) )
2009-05-27 15:16:21 +02:00
goto done ;
2007-07-12 10:49:39 +02:00
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
/* Dump triggers. */
2012-01-10 09:03:45 +01:00
if ( ! mysql_num_rows ( show_triggers_rs ) )
goto skip ;
if ( opt_xml )
print_xml_tag ( sql_file , " \t " , " \n " , " triggers " , " name= " ,
table_name , NullS ) ;
2007-08-03 09:50:17 +02:00
while ( ( row = mysql_fetch_row ( show_triggers_rs ) ) )
2005-09-09 00:03:07 +02:00
{
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
my_snprintf ( query_buff , sizeof ( query_buff ) ,
" SHOW CREATE TRIGGER %s " ,
quote_name ( row [ 0 ] , name_buff , TRUE ) ) ;
2006-01-11 00:07:40 +01:00
2007-08-03 09:50:17 +02:00
if ( mysql_query ( mysql , query_buff ) )
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
{
2006-01-11 00:07:40 +01:00
/*
2007-08-03 09:50:17 +02:00
mysqldump is being run against old server , that does not support
SHOW CREATE TRIGGER statement . We should use SHOW TRIGGERS output .
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-08-03 09:50:17 +02:00
NOTE : the dump may be incorrect , as old SHOW TRIGGERS does not
provide all the necessary information to restore trigger properly .
*/
2006-01-11 00:07:40 +01:00
2009-05-27 15:16:21 +02:00
dump_trigger_old ( sql_file , show_triggers_rs , & row , table_name ) ;
2007-08-03 09:50:17 +02:00
}
else
{
MYSQL_RES * show_create_trigger_rs = mysql_store_result ( mysql ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-08-03 09:50:17 +02:00
if ( ! show_create_trigger_rs | |
2009-05-27 15:16:21 +02:00
dump_trigger ( sql_file , show_create_trigger_rs , db_name , db_cl_name ) )
goto done ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-08-03 09:50:17 +02:00
mysql_free_result ( show_create_trigger_rs ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
}
2007-08-03 09:50:17 +02:00
2005-09-09 00:03:07 +02:00
}
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2012-01-10 09:03:45 +01:00
if ( opt_xml )
{
fputs ( " \t </triggers> \n " , sql_file ) ;
check_io ( sql_file ) ;
}
skip :
2007-08-03 09:50:17 +02:00
mysql_free_result ( show_triggers_rs ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2007-07-12 10:49:39 +02:00
if ( switch_character_set_results ( mysql , default_charset ) )
2009-05-27 15:16:21 +02:00
goto done ;
2007-07-12 10:49:39 +02:00
2005-10-13 22:42:56 +02:00
/*
2006-05-25 10:00:28 +02:00
make sure to set back opt_compatible mode to
2005-10-13 22:42:56 +02:00
original value
*/
opt_compatible_mode = old_opt_compatible_mode ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2009-05-27 15:16:21 +02:00
ret = FALSE ;
done :
if ( path )
my_fclose ( sql_file , MYF ( 0 ) ) ;
DBUG_RETURN ( ret ) ;
2005-09-09 00:03:07 +02:00
}
2007-03-22 18:35:29 +01:00
static void add_load_option ( DYNAMIC_STRING * str , const char * option ,
const char * option_value )
2000-07-31 21:29:14 +02:00
{
2007-03-22 18:35:29 +01:00
if ( ! option_value )
2000-07-31 21:29:14 +02:00
{
2007-03-22 18:35:29 +01:00
/* Null value means we don't add this option. */
return ;
2000-07-31 21:29:14 +02:00
}
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( str , option ) ;
if ( strncmp ( option_value , " 0x " , sizeof ( " 0x " ) - 1 ) = = 0 )
{
/* It's a hex constant, don't escape */
dynstr_append_checked ( str , option_value ) ;
2000-07-31 21:29:14 +02:00
}
2007-03-22 18:35:29 +01:00
else
{
/* char constant; escape */
field_escape ( str , option_value ) ;
}
}
2000-07-31 21:29:14 +02:00
/*
2005-10-06 16:54:43 +02:00
Allow the user to specify field terminator strings like :
" ' " , " \" , " \ \ " (escaped backslash), " \ t " (tab), " \ n " (newline)
This is done by doubling ' and add a end - \ if needed to avoid
syntax errors from the SQL parser .
2000-07-31 21:29:14 +02:00
*/
2007-03-22 18:35:29 +01:00
static void field_escape ( DYNAMIC_STRING * in , const char * from )
2000-07-31 21:29:14 +02:00
{
2007-03-22 18:35:29 +01:00
uint end_backslashes = 0 ;
2000-07-31 21:29:14 +02:00
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( in , " ' " ) ;
2000-07-31 21:29:14 +02:00
2007-03-22 18:35:29 +01:00
while ( * from )
2000-07-31 21:29:14 +02:00
{
2007-03-22 18:35:29 +01:00
dynstr_append_mem_checked ( in , from , 1 ) ;
2000-07-31 21:29:14 +02:00
if ( * from = = ' \\ ' )
end_backslashes ^ = 1 ; /* find odd number of backslashes */
else
{
if ( * from = = ' \' ' & & ! end_backslashes )
2007-03-22 18:35:29 +01:00
{
/* We want a duplicate of "'" for MySQL */
dynstr_append_checked ( in , " \' " ) ;
}
2000-07-31 21:29:14 +02:00
end_backslashes = 0 ;
}
2007-03-22 18:35:29 +01:00
from + + ;
2000-07-31 21:29:14 +02:00
}
/* Add missing backslashes if user has specified odd number of backs.*/
if ( end_backslashes )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( in , " \\ " ) ;
dynstr_append_checked ( in , " ' " ) ;
}
2000-07-31 21:29:14 +02:00
2004-05-07 00:02:57 +02:00
static char * alloc_query_str ( ulong size )
{
char * query ;
if ( ! ( query = ( char * ) my_malloc ( size , MYF ( MY_WME ) ) ) )
2007-03-22 18:35:29 +01:00
die ( EX_MYSQLERR , " Couldn't allocate a query string. " ) ;
2004-05-07 00:02:57 +02:00
return query ;
}
2005-04-04 23:32:48 +02:00
2000-07-31 21:29:14 +02:00
/*
2005-09-27 07:43:09 +02:00
SYNOPSIS
dump_table ( )
dump_table saves database contents as a series of INSERT statements .
ARGS
table - table name
db - db name
RETURNS
void
2000-07-31 21:29:14 +02:00
*/
2005-04-04 23:32:48 +02:00
2007-03-22 18:35:29 +01:00
2005-09-27 07:43:09 +02:00
static void dump_table ( char * table , char * db )
2000-07-31 21:29:14 +02:00
{
2005-09-27 07:43:09 +02:00
char ignore_flag ;
2007-03-22 18:35:29 +01:00
char buf [ 200 ] , table_buff [ NAME_LEN + 3 ] ;
DYNAMIC_STRING query_string ;
2005-10-11 23:58:22 +02:00
char table_type [ NAME_LEN ] ;
2003-09-05 05:56:28 +02:00
char * result_table , table_buff2 [ NAME_LEN * 2 + 3 ] , * opt_quoted_table ;
2005-09-27 07:43:09 +02:00
int error = 0 ;
2006-05-25 10:00:28 +02:00
ulong rownr , row_break , total_length , init_length ;
2005-09-27 07:43:09 +02:00
uint num_fields ;
2006-05-25 10:00:28 +02:00
MYSQL_RES * res ;
MYSQL_FIELD * field ;
MYSQL_ROW row ;
2005-09-27 07:43:09 +02:00
DBUG_ENTER ( " dump_table " ) ;
/*
Make sure you get the create table info before the following check for
- - no - data flag below . Otherwise , the create table info won ' t be printed .
*/
2005-10-11 23:58:22 +02:00
num_fields = get_table_structure ( table , db , table_type , & ignore_flag ) ;
2000-07-31 21:29:14 +02:00
2006-05-25 10:00:28 +02:00
/*
The " table " could be a view . If so , we don ' t do anything here .
*/
2007-08-02 06:49:29 +02:00
if ( strcmp ( table_type , " VIEW " ) = = 0 )
2006-11-27 00:47:38 +01:00
DBUG_VOID_RETURN ;
2006-05-25 10:00:28 +02:00
2005-06-22 15:43:49 +02:00
/* Check --no-data flag */
2006-07-24 11:56:01 +02:00
if ( opt_no_data )
2005-06-22 15:43:49 +02:00
{
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Skipping dump data for table '%s', --no-data was used \n " ,
table ) ;
2005-09-27 07:43:09 +02:00
DBUG_VOID_RETURN ;
2005-06-22 15:43:49 +02:00
}
2005-10-11 23:58:22 +02:00
DBUG_PRINT ( " info " ,
( " ignore_flag: %x num_fields: %d " , ( int ) ignore_flag ,
num_fields ) ) ;
2005-09-27 07:43:09 +02:00
/*
If the table type is a merge table or any type that has to be
_completely_ ignored and no data dumped
*/
if ( ignore_flag & IGNORE_DATA )
{
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Warning: Skipping data for table '%s' because " \
" it's of type %s \n " , table , table_type ) ;
2005-09-27 07:43:09 +02:00
DBUG_VOID_RETURN ;
}
2005-06-22 15:43:49 +02:00
/* Check that there are any fields in the table */
2005-09-27 07:43:09 +02:00
if ( num_fields = = 0 )
2005-06-22 15:43:49 +02:00
{
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Skipping dump data for table '%s', it has no fields \n " ,
table ) ;
2005-09-27 07:43:09 +02:00
DBUG_VOID_RETURN ;
2005-06-22 15:43:49 +02:00
}
2007-09-05 08:35:29 +02:00
/*
Check - - skip - events flag : it is not enough to skip creation of events
discarding SHOW CREATE EVENT statements generation . The myslq . event
table data should be skipped too .
*/
if ( ! opt_events & & ! my_strcasecmp ( & my_charset_latin1 , db , " mysql " ) & &
! my_strcasecmp ( & my_charset_latin1 , table , " event " ) )
{
2012-11-09 10:45:16 +01:00
fprintf ( stderr , " -- Warning: Skipping the data of table mysql.event. "
" Specify the --events option explicitly. \n " ) ;
2007-09-05 08:35:29 +02:00
DBUG_VOID_RETURN ;
}
2003-09-03 17:48:10 +02:00
result_table = quote_name ( table , table_buff , 1 ) ;
opt_quoted_table = quote_name ( table , table_buff2 , 0 ) ;
2003-12-14 05:39:52 +01:00
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Sending SELECT query... \n " ) ;
2007-03-22 18:35:29 +01:00
init_dynamic_string_checked ( & query_string , " " , 1024 , 1024 ) ;
2000-07-31 21:29:14 +02:00
if ( path )
{
char filename [ FN_REFLEN ] , tmp_path [ FN_REFLEN ] ;
2007-03-22 18:35:29 +01:00
/*
Convert the path to native os format
and resolve to the full filepath .
*/
convert_dirname ( tmp_path , path , NullS ) ;
2000-07-31 21:29:14 +02:00
my_load_path ( tmp_path , tmp_path , NULL ) ;
2007-03-22 18:35:29 +01:00
fn_format ( filename , table , tmp_path , " .txt " , MYF ( MY_UNPACK_FILENAME ) ) ;
/* Must delete the file that 'INTO OUTFILE' will write to */
my_delete ( filename , MYF ( 0 ) ) ;
/* convert to a unix path name to stick into the query */
2000-07-31 21:29:14 +02:00
to_unix_path ( filename ) ;
2007-03-22 18:35:29 +01:00
/* now build the query string */
dynstr_append_checked ( & query_string , " SELECT /*!40001 SQL_NO_CACHE */ * INTO OUTFILE ' " ) ;
dynstr_append_checked ( & query_string , filename ) ;
dynstr_append_checked ( & query_string , " ' " ) ;
2000-07-31 21:29:14 +02:00
2009-08-03 21:22:28 +02:00
dynstr_append_checked ( & query_string , " /*!50138 CHARACTER SET " ) ;
2009-07-31 19:14:52 +02:00
dynstr_append_checked ( & query_string , default_charset = = mysql_universal_client_charset ?
my_charset_bin . name : /* backward compatibility */
default_charset ) ;
dynstr_append_checked ( & query_string , " */ " ) ;
2000-07-31 21:29:14 +02:00
if ( fields_terminated | | enclosed | | opt_enclosed | | escaped )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & query_string , " FIELDS " ) ;
add_load_option ( & query_string , " TERMINATED BY " , fields_terminated ) ;
add_load_option ( & query_string , " ENCLOSED BY " , enclosed ) ;
add_load_option ( & query_string , " OPTIONALLY ENCLOSED BY " , opt_enclosed ) ;
add_load_option ( & query_string , " ESCAPED BY " , escaped ) ;
add_load_option ( & query_string , " LINES TERMINATED BY " , lines_terminated ) ;
dynstr_append_checked ( & query_string , " FROM " ) ;
dynstr_append_checked ( & query_string , result_table ) ;
if ( where )
2004-05-07 00:02:57 +02:00
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & query_string , " WHERE " ) ;
dynstr_append_checked ( & query_string , where ) ;
2004-05-07 00:02:57 +02:00
}
2007-03-22 18:35:29 +01:00
if ( order_by )
{
dynstr_append_checked ( & query_string , " ORDER BY " ) ;
dynstr_append_checked ( & query_string , order_by ) ;
}
if ( mysql_real_query ( mysql , query_string . str , query_string . length ) )
2000-07-31 21:29:14 +02:00
{
2006-07-24 12:00:44 +02:00
DB_error ( mysql , " when executing 'SELECT INTO OUTFILE' " ) ;
2007-08-02 06:49:29 +02:00
dynstr_free ( & query_string ) ;
2005-09-27 07:43:09 +02:00
DBUG_VOID_RETURN ;
2000-07-31 21:29:14 +02:00
}
}
else
{
2012-01-10 09:03:45 +01:00
print_comment ( md_result_file , 0 ,
" \n -- \n -- Dumping data for table %s \n -- \n " ,
result_table ) ;
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & query_string , " SELECT /*!40001 SQL_NO_CACHE */ * FROM " ) ;
dynstr_append_checked ( & query_string , result_table ) ;
2004-11-16 23:45:24 +01:00
2007-03-22 18:35:29 +01:00
if ( where )
{
2012-01-10 09:03:45 +01:00
print_comment ( md_result_file , 0 , " -- WHERE: %s \n " , where ) ;
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & query_string , " WHERE " ) ;
dynstr_append_checked ( & query_string , where ) ;
}
if ( order_by )
{
2012-01-10 09:03:45 +01:00
print_comment ( md_result_file , 0 , " -- ORDER BY: %s \n " , order_by ) ;
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & query_string , " ORDER BY " ) ;
dynstr_append_checked ( & query_string , order_by ) ;
2000-07-31 21:29:14 +02:00
}
2007-03-22 18:35:29 +01:00
2004-02-09 12:31:03 +01:00
if ( ! opt_xml & & ! opt_compact )
2004-06-24 22:14:12 +02:00
{
2002-04-03 12:36:01 +02:00
fputs ( " \n " , md_result_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
}
2007-03-22 18:35:29 +01:00
if ( mysql_query_with_error_report ( mysql , 0 , query_string . str ) )
2006-08-17 23:09:24 +02:00
{
2006-07-24 12:00:44 +02:00
DB_error ( mysql , " when retrieving data from server " ) ;
2006-08-17 23:09:24 +02:00
goto err ;
}
2000-07-31 21:29:14 +02:00
if ( quick )
2006-07-24 12:00:44 +02:00
res = mysql_use_result ( mysql ) ;
2000-07-31 21:29:14 +02:00
else
2006-07-24 12:00:44 +02:00
res = mysql_store_result ( mysql ) ;
2000-07-31 21:29:14 +02:00
if ( ! res )
2006-01-03 17:54:54 +01:00
{
2006-07-24 12:00:44 +02:00
DB_error ( mysql , " when retrieving data from server " ) ;
2006-01-03 17:54:54 +01:00
goto err ;
}
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Retrieving rows... \n " ) ;
2005-09-27 07:43:09 +02:00
if ( mysql_num_fields ( res ) ! = num_fields )
2000-07-31 21:29:14 +02:00
{
2003-09-03 17:48:10 +02:00
fprintf ( stderr , " %s: Error in field count for table: %s ! Aborting. \n " ,
2006-05-25 10:00:28 +02:00
my_progname , result_table ) ;
2004-05-07 00:02:57 +02:00
error = EX_CONSCHECK ;
goto err ;
2000-07-31 21:29:14 +02:00
}
2006-06-21 09:12:23 +02:00
if ( opt_lock )
2004-06-24 22:14:12 +02:00
{
2006-06-21 09:12:23 +02:00
fprintf ( md_result_file , " LOCK TABLES %s WRITE; \n " , opt_quoted_table ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
}
2006-06-21 09:12:23 +02:00
/* Moved disable keys to after lock per bug 15977 */
if ( opt_disable_keys )
2004-06-24 22:14:12 +02:00
{
2006-06-21 09:12:23 +02:00
fprintf ( md_result_file , " /*!40000 ALTER TABLE %s DISABLE KEYS */; \n " ,
opt_quoted_table ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
}
2000-07-31 21:29:14 +02:00
2006-05-25 10:00:28 +02:00
total_length = opt_net_buffer_length ; /* Force row break */
2000-07-31 21:29:14 +02:00
row_break = 0 ;
rownr = 0 ;
2005-05-08 21:02:46 +02:00
init_length = ( uint ) insert_pat . length + 4 ;
2001-11-05 22:48:03 +01:00
if ( opt_xml )
2006-10-19 00:43:51 +02:00
print_xml_tag ( md_result_file , " \t " , " \n " , " table_data " , " name= " , table ,
NullS ) ;
2001-10-04 19:43:06 +02:00
if ( opt_autocommit )
2004-06-24 22:14:12 +02:00
{
2001-10-04 19:43:06 +02:00
fprintf ( md_result_file , " set autocommit=0; \n " ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
}
2001-10-04 19:43:06 +02:00
2005-10-25 19:04:31 +02:00
while ( ( row = mysql_fetch_row ( res ) ) )
2000-07-31 21:29:14 +02:00
{
uint i ;
2005-10-25 19:04:31 +02:00
ulong * lengths = mysql_fetch_lengths ( res ) ;
2000-07-31 21:29:14 +02:00
rownr + + ;
2001-11-05 22:48:03 +01:00
if ( ! extended_insert & & ! opt_xml )
2004-06-24 22:14:12 +02:00
{
2006-05-25 10:00:28 +02:00
fputs ( insert_pat . str , md_result_file ) ;
check_io ( md_result_file ) ;
2004-06-24 22:14:12 +02:00
}
2000-07-31 21:29:14 +02:00
mysql_field_seek ( res , 0 ) ;
2002-04-03 12:36:01 +02:00
if ( opt_xml )
2004-06-24 22:14:12 +02:00
{
2003-10-30 11:58:30 +01:00
fputs ( " \t <row> \n " , md_result_file ) ;
2006-05-25 10:00:28 +02:00
check_io ( md_result_file ) ;
2004-06-24 22:14:12 +02:00
}
2002-04-03 12:36:01 +02:00
2006-08-02 03:18:47 +02:00
for ( i = 0 ; i < mysql_num_fields ( res ) ; i + + )
2000-07-31 21:29:14 +02:00
{
2004-10-19 07:40:59 +02:00
int is_blob ;
2005-11-24 02:31:44 +01:00
ulong length = lengths [ i ] ;
2006-08-02 03:18:47 +02:00
if ( ! ( field = mysql_fetch_field ( res ) ) )
2007-03-22 18:35:29 +01:00
die ( EX_CONSCHECK ,
" Not enough fields from table %s! Aborting. \n " ,
result_table ) ;
2006-05-25 10:00:28 +02:00
/*
63 is my_charset_bin . If charsetnr is not 63 ,
we have not a BLOB but a TEXT column .
we ' ll dump in hex only BLOB columns .
*/
2004-10-19 07:40:59 +02:00
is_blob = ( opt_hex_blob & & field - > charsetnr = = 63 & &
2005-09-14 15:46:31 +02:00
( field - > type = = MYSQL_TYPE_BIT | |
field - > type = = MYSQL_TYPE_STRING | |
2004-12-06 01:00:37 +01:00
field - > type = = MYSQL_TYPE_VAR_STRING | |
field - > type = = MYSQL_TYPE_VARCHAR | |
field - > type = = MYSQL_TYPE_BLOB | |
field - > type = = MYSQL_TYPE_LONG_BLOB | |
field - > type = = MYSQL_TYPE_MEDIUM_BLOB | |
field - > type = = MYSQL_TYPE_TINY_BLOB ) ) ? 1 : 0 ;
2006-10-19 00:43:51 +02:00
if ( extended_insert & & ! opt_xml )
2006-05-25 10:00:28 +02:00
{
if ( i = = 0 )
2007-03-22 18:35:29 +01:00
dynstr_set_checked ( & extended_row , " ( " ) ;
2006-05-25 10:00:28 +02:00
else
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , " , " ) ;
2006-05-25 10:00:28 +02:00
if ( row [ i ] )
{
if ( length )
{
2009-09-30 12:25:50 +02:00
if ( ! ( field - > flags & NUM_FLAG ) )
2006-05-25 10:00:28 +02:00
{
/*
" length * 2 + 2 " is OK for both HEX and non - HEX modes :
- In HEX mode we need exactly 2 bytes per character
plus 2 bytes for ' 0 x ' prefix .
- In non - HEX mode we need up to 2 bytes per character ,
plus 2 bytes for leading and trailing ' \' ' characters .
2007-05-25 14:24:17 +02:00
Also we need to reserve 1 byte for terminating ' \0 ' .
2006-05-25 10:00:28 +02:00
*/
2007-05-25 14:24:17 +02:00
dynstr_realloc_checked ( & extended_row , length * 2 + 2 + 1 ) ;
2004-10-19 07:40:59 +02:00
if ( opt_hex_blob & & is_blob )
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , " 0x " ) ;
2006-05-25 10:00:28 +02:00
extended_row . length + = mysql_hex_string ( extended_row . str +
2004-11-05 05:54:52 +01:00
extended_row . length ,
row [ i ] , length ) ;
2007-05-25 14:24:17 +02:00
DBUG_ASSERT ( extended_row . length + 1 < = extended_row . max_length ) ;
/* mysql_hex_string() already terminated string by '\0' */
DBUG_ASSERT ( extended_row . str [ extended_row . length ] = = ' \0 ' ) ;
2004-10-19 07:40:59 +02:00
}
else
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , " ' " ) ;
2004-10-19 07:40:59 +02:00
extended_row . length + =
mysql_real_escape_string ( & mysql_connection ,
& extended_row . str [ extended_row . length ] ,
row [ i ] , length ) ;
extended_row . str [ extended_row . length ] = ' \0 ' ;
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , " ' " ) ;
2004-10-19 07:40:59 +02:00
}
2006-05-25 10:00:28 +02:00
}
else
{
/* change any strings ("inf", "-inf", "nan") into NULL */
2006-08-02 03:18:47 +02:00
char * ptr = row [ i ] ;
2006-05-25 10:00:28 +02:00
if ( my_isalpha ( charset_info , * ptr ) | | ( * ptr = = ' - ' & &
my_isalpha ( charset_info , ptr [ 1 ] ) ) )
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , " NULL " ) ;
2006-05-25 10:00:28 +02:00
else
{
2006-12-02 02:26:52 +01:00
if ( field - > type = = MYSQL_TYPE_DECIMAL )
2006-05-25 10:00:28 +02:00
{
/* add " signs around */
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , " ' " ) ;
dynstr_append_checked ( & extended_row , ptr ) ;
dynstr_append_checked ( & extended_row , " ' " ) ;
2006-05-25 10:00:28 +02:00
}
else
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , ptr ) ;
2006-05-25 10:00:28 +02:00
}
}
}
else
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , " '' " ) ;
2006-05-25 10:00:28 +02:00
}
2007-03-22 18:35:29 +01:00
else
dynstr_append_checked ( & extended_row , " NULL " ) ;
2006-05-25 10:00:28 +02:00
}
else
{
if ( i & & ! opt_xml )
{
fputc ( ' , ' , md_result_file ) ;
check_io ( md_result_file ) ;
}
if ( row [ i ] )
{
2009-09-30 12:25:50 +02:00
if ( ! ( field - > flags & NUM_FLAG ) )
2006-05-25 10:00:28 +02:00
{
if ( opt_xml )
{
2006-10-19 00:43:51 +02:00
if ( opt_hex_blob & & is_blob & & length )
{
2006-11-27 00:47:38 +01:00
/* Define xsi:type="xs:hexBinary" for hex encoded data */
print_xml_tag ( md_result_file , " \t \t " , " " , " field " , " name= " ,
field - > name , " xsi:type= " , " xs:hexBinary " , NullS ) ;
print_blob_as_hex ( md_result_file , row [ i ] , length ) ;
2006-10-19 00:43:51 +02:00
}
else
{
2006-11-27 00:47:38 +01:00
print_xml_tag ( md_result_file , " \t \t " , " " , " field " , " name= " ,
field - > name , NullS ) ;
2012-01-10 09:03:45 +01:00
print_quoted_xml ( md_result_file , row [ i ] , length , 0 ) ;
2006-10-19 00:43:51 +02:00
}
2006-05-25 10:00:28 +02:00
fputs ( " </field> \n " , md_result_file ) ;
}
else if ( opt_hex_blob & & is_blob & & length )
2004-11-12 10:17:53 +01:00
{
2004-10-19 07:40:59 +02:00
fputs ( " 0x " , md_result_file ) ;
2006-10-19 00:43:51 +02:00
print_blob_as_hex ( md_result_file , row [ i ] , length ) ;
2004-10-19 07:40:59 +02:00
}
else
2005-11-24 02:31:44 +01:00
unescape ( md_result_file , row [ i ] , length ) ;
2006-05-25 10:00:28 +02:00
}
else
{
/* change any strings ("inf", "-inf", "nan") into NULL */
2006-08-02 03:18:47 +02:00
char * ptr = row [ i ] ;
2006-05-25 10:00:28 +02:00
if ( opt_xml )
{
2006-10-19 00:43:51 +02:00
print_xml_tag ( md_result_file , " \t \t " , " " , " field " , " name= " ,
field - > name , NullS ) ;
2006-05-25 10:00:28 +02:00
fputs ( ! my_isalpha ( charset_info , * ptr ) ? ptr : " NULL " ,
md_result_file ) ;
fputs ( " </field> \n " , md_result_file ) ;
}
else if ( my_isalpha ( charset_info , * ptr ) | |
( * ptr = = ' - ' & & my_isalpha ( charset_info , ptr [ 1 ] ) ) )
fputs ( " NULL " , md_result_file ) ;
2006-12-02 02:26:52 +01:00
else if ( field - > type = = MYSQL_TYPE_DECIMAL )
2006-05-25 10:00:28 +02:00
{
/* add " signs around */
fputc ( ' \' ' , md_result_file ) ;
fputs ( ptr , md_result_file ) ;
fputc ( ' \' ' , md_result_file ) ;
}
else
fputs ( ptr , md_result_file ) ;
}
}
else
2005-06-22 20:22:54 +02:00
{
/* The field value is NULL */
if ( ! opt_xml )
fputs ( " NULL " , md_result_file ) ;
else
print_xml_null_tag ( md_result_file , " \t \t " , " field name= " ,
field - > name , " \n " ) ;
}
2004-06-24 23:29:27 +02:00
check_io ( md_result_file ) ;
2006-05-25 10:00:28 +02:00
}
2000-07-31 21:29:14 +02:00
}
2002-04-03 12:36:01 +02:00
if ( opt_xml )
2004-06-24 22:14:12 +02:00
{
2003-10-30 11:58:30 +01:00
fputs ( " \t </row> \n " , md_result_file ) ;
2006-05-25 10:00:28 +02:00
check_io ( md_result_file ) ;
2004-06-24 22:14:12 +02:00
}
2002-04-03 12:36:01 +02:00
2003-10-14 12:19:54 +02:00
if ( extended_insert )
2000-07-31 21:29:14 +02:00
{
2006-05-25 10:00:28 +02:00
ulong row_length ;
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & extended_row , " ) " ) ;
2006-08-02 03:18:47 +02:00
row_length = 2 + extended_row . length ;
2004-05-27 01:47:04 +02:00
if ( total_length + row_length < opt_net_buffer_length )
2000-07-31 21:29:14 +02:00
{
2006-08-02 03:18:47 +02:00
total_length + = row_length ;
2006-05-25 10:00:28 +02:00
fputc ( ' , ' , md_result_file ) ; /* Always row break */
fputs ( extended_row . str , md_result_file ) ;
}
2000-07-31 21:29:14 +02:00
else
{
2006-05-25 10:00:28 +02:00
if ( row_break )
fputs ( " ; \n " , md_result_file ) ;
row_break = 1 ; /* This is first row */
2001-12-06 13:10:51 +01:00
2005-05-08 21:02:46 +02:00
fputs ( insert_pat . str , md_result_file ) ;
2003-10-14 12:19:54 +02:00
fputs ( extended_row . str , md_result_file ) ;
2006-08-02 03:18:47 +02:00
total_length = row_length + init_length ;
2000-07-31 21:29:14 +02:00
}
2006-05-25 10:00:28 +02:00
check_io ( md_result_file ) ;
2000-07-31 21:29:14 +02:00
}
2001-11-05 22:48:03 +01:00
else if ( ! opt_xml )
2004-06-24 22:14:12 +02:00
{
2006-05-25 10:00:28 +02:00
fputs ( " ); \n " , md_result_file ) ;
check_io ( md_result_file ) ;
2004-06-24 22:14:12 +02:00
}
2000-07-31 21:29:14 +02:00
}
2001-12-06 13:10:51 +01:00
2002-01-22 23:04:43 +01:00
/* XML - close table tag and supress regular output */
2001-11-05 22:48:03 +01:00
if ( opt_xml )
2006-05-25 10:00:28 +02:00
fputs ( " \t </table_data> \n " , md_result_file ) ;
2001-11-05 22:48:03 +01:00
else if ( extended_insert & & row_break )
2006-05-25 10:00:28 +02:00
fputs ( " ; \n " , md_result_file ) ; /* If not empty table */
2001-08-22 00:45:07 +02:00
fflush ( md_result_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
2006-07-24 12:00:44 +02:00
if ( mysql_errno ( mysql ) )
2000-07-31 21:29:14 +02:00
{
2007-03-22 18:35:29 +01:00
my_snprintf ( buf , sizeof ( buf ) ,
2006-05-25 10:00:28 +02:00
" %s: Error %d: %s when dumping table %s at row: %ld \n " ,
my_progname ,
2006-07-24 12:00:44 +02:00
mysql_errno ( mysql ) ,
mysql_error ( mysql ) ,
2006-05-25 10:00:28 +02:00
result_table ,
rownr ) ;
2007-03-22 18:35:29 +01:00
fputs ( buf , stderr ) ;
2004-05-07 00:02:57 +02:00
error = EX_CONSCHECK ;
goto err ;
2000-07-31 21:29:14 +02:00
}
2006-06-21 09:12:23 +02:00
/* Moved enable keys to before unlock per bug 15977 */
2002-04-03 12:36:01 +02:00
if ( opt_disable_keys )
2004-06-24 22:14:12 +02:00
{
2002-04-03 12:36:01 +02:00
fprintf ( md_result_file , " /*!40000 ALTER TABLE %s ENABLE KEYS */; \n " ,
2006-05-25 10:00:28 +02:00
opt_quoted_table ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
}
2006-06-21 09:12:23 +02:00
if ( opt_lock )
{
fputs ( " UNLOCK TABLES; \n " , md_result_file ) ;
check_io ( md_result_file ) ;
}
2001-10-04 19:43:06 +02:00
if ( opt_autocommit )
2004-06-24 22:14:12 +02:00
{
2001-10-04 19:43:06 +02:00
fprintf ( md_result_file , " commit; \n " ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
}
2001-10-10 22:44:07 +02:00
mysql_free_result ( res ) ;
2005-09-27 07:43:09 +02:00
}
2007-08-02 06:49:29 +02:00
dynstr_free ( & query_string ) ;
2005-09-27 07:43:09 +02:00
DBUG_VOID_RETURN ;
2004-05-07 00:02:57 +02:00
err :
2007-03-22 18:35:29 +01:00
dynstr_free ( & query_string ) ;
maybe_exit ( error ) ;
2005-09-27 07:43:09 +02:00
DBUG_VOID_RETURN ;
2005-04-04 23:32:48 +02:00
} /* dump_table */
2000-07-31 21:29:14 +02:00
static char * getTableName ( int reset )
{
2006-08-02 03:18:47 +02:00
static MYSQL_RES * res = NULL ;
2000-07-31 21:29:14 +02:00
MYSQL_ROW row ;
if ( ! res )
{
2006-09-11 17:14:40 +02:00
if ( ! ( res = mysql_list_tables ( mysql , NullS ) ) )
2000-07-31 21:29:14 +02:00
return ( NULL ) ;
}
2006-08-02 03:18:47 +02:00
if ( ( row = mysql_fetch_row ( res ) ) )
2000-07-31 21:29:14 +02:00
return ( ( char * ) row [ 0 ] ) ;
2000-09-07 03:55:17 +02:00
2000-07-31 21:29:14 +02:00
if ( reset )
mysql_data_seek ( res , 0 ) ; /* We want to read again */
else
{
mysql_free_result ( res ) ;
2006-08-02 03:18:47 +02:00
res = NULL ;
2000-07-31 21:29:14 +02:00
}
return ( NULL ) ;
} /* getTableName */
2006-02-02 00:52:58 +01:00
/*
dump all logfile groups and tablespaces
*/
static int dump_all_tablespaces ( )
2006-10-09 16:03:46 +02:00
{
return dump_tablespaces ( NULL ) ;
}
static int dump_tablespaces_for_tables ( char * db , char * * table_names , int tables )
{
2006-10-16 17:02:12 +02:00
DYNAMIC_STRING where ;
2006-10-09 16:03:46 +02:00
int r ;
int i ;
char name_buff [ NAME_LEN * 2 + 3 ] ;
2006-10-16 17:02:12 +02:00
2006-10-09 16:03:46 +02:00
mysql_real_escape_string ( mysql , name_buff , db , strlen ( db ) ) ;
2007-03-23 01:36:58 +01:00
init_dynamic_string_checked ( & where , " AND TABLESPACE_NAME IN ( "
2006-10-16 17:02:12 +02:00
" SELECT DISTINCT TABLESPACE_NAME FROM "
" INFORMATION_SCHEMA.PARTITIONS "
" WHERE "
" TABLE_SCHEMA=' " , 256 , 1024 ) ;
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & where , name_buff ) ;
dynstr_append_checked ( & where , " ' AND TABLE_NAME IN ( " ) ;
2006-10-09 16:03:46 +02:00
for ( i = 0 ; i < tables ; i + + )
{
mysql_real_escape_string ( mysql , name_buff ,
table_names [ i ] , strlen ( table_names [ i ] ) ) ;
2006-10-16 17:02:12 +02:00
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & where , " ' " ) ;
dynstr_append_checked ( & where , name_buff ) ;
dynstr_append_checked ( & where , " ', " ) ;
2006-10-09 16:03:46 +02:00
}
2006-10-20 14:47:28 +02:00
dynstr_trunc ( & where , 1 ) ;
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & where , " )) " ) ;
2006-10-09 16:03:46 +02:00
2007-01-29 00:47:35 +01:00
DBUG_PRINT ( " info " , ( " Dump TS for Tables where: %s " , where . str ) ) ;
2006-10-16 17:02:12 +02:00
r = dump_tablespaces ( where . str ) ;
dynstr_free ( & where ) ;
2006-10-09 16:03:46 +02:00
return r ;
}
static int dump_tablespaces_for_databases ( char * * databases )
{
2006-10-16 17:02:12 +02:00
DYNAMIC_STRING where ;
2006-10-09 16:03:46 +02:00
int r ;
int i ;
2007-03-23 01:36:58 +01:00
init_dynamic_string_checked ( & where , " AND TABLESPACE_NAME IN ( "
2006-10-16 17:02:12 +02:00
" SELECT DISTINCT TABLESPACE_NAME FROM "
" INFORMATION_SCHEMA.PARTITIONS "
" WHERE "
" TABLE_SCHEMA IN ( " , 256 , 1024 ) ;
2006-10-09 16:03:46 +02:00
for ( i = 0 ; databases [ i ] ! = NULL ; i + + )
{
char db_name_buff [ NAME_LEN * 2 + 3 ] ;
mysql_real_escape_string ( mysql , db_name_buff ,
databases [ i ] , strlen ( databases [ i ] ) ) ;
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & where , " ' " ) ;
dynstr_append_checked ( & where , db_name_buff ) ;
dynstr_append_checked ( & where , " ', " ) ;
2006-10-09 16:03:46 +02:00
}
2006-10-20 14:47:28 +02:00
dynstr_trunc ( & where , 1 ) ;
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & where , " )) " ) ;
2006-10-09 16:03:46 +02:00
2007-01-29 00:47:35 +01:00
DBUG_PRINT ( " info " , ( " Dump TS for DBs where: %s " , where . str ) ) ;
2006-10-16 17:02:12 +02:00
r = dump_tablespaces ( where . str ) ;
dynstr_free ( & where ) ;
2006-10-09 16:03:46 +02:00
return r ;
}
static int dump_tablespaces ( char * ts_where )
2006-02-02 00:52:58 +01:00
{
MYSQL_ROW row ;
MYSQL_RES * tableres ;
char buf [ FN_REFLEN ] ;
2006-10-16 17:02:12 +02:00
DYNAMIC_STRING sqlbuf ;
2007-01-29 00:47:35 +01:00
int first = 0 ;
2006-09-19 09:07:21 +02:00
/*
The following are used for parsing the EXTRA field
*/
char extra_format [ ] = " UNDO_BUFFER_SIZE= " ;
char * ubs ;
char * endsemi ;
2007-08-02 06:49:29 +02:00
DBUG_ENTER ( " dump_tablespaces " ) ;
2006-02-02 00:52:58 +01:00
2007-03-23 01:36:58 +01:00
init_dynamic_string_checked ( & sqlbuf ,
2006-10-16 17:02:12 +02:00
" SELECT LOGFILE_GROUP_NAME, "
" FILE_NAME, "
" TOTAL_EXTENTS, "
" INITIAL_SIZE, "
" ENGINE, "
" EXTRA "
" FROM INFORMATION_SCHEMA.FILES "
" WHERE FILE_TYPE = 'UNDO LOG' "
" AND FILE_NAME IS NOT NULL " ,
256 , 1024 ) ;
if ( ts_where )
{
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & sqlbuf ,
2006-10-16 17:02:12 +02:00
" AND LOGFILE_GROUP_NAME IN ( "
" SELECT DISTINCT LOGFILE_GROUP_NAME "
" FROM INFORMATION_SCHEMA.FILES "
" WHERE FILE_TYPE = 'DATAFILE' "
) ;
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & sqlbuf , ts_where ) ;
dynstr_append_checked ( & sqlbuf , " ) " ) ;
2006-10-16 17:02:12 +02:00
}
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & sqlbuf ,
2006-10-16 17:02:12 +02:00
" GROUP BY LOGFILE_GROUP_NAME, FILE_NAME "
" , ENGINE "
" ORDER BY LOGFILE_GROUP_NAME " ) ;
if ( mysql_query ( mysql , sqlbuf . str ) | |
2006-10-16 16:15:58 +02:00
! ( tableres = mysql_store_result ( mysql ) ) )
{
2007-08-02 06:49:29 +02:00
dynstr_free ( & sqlbuf ) ;
2006-10-16 16:15:58 +02:00
if ( mysql_errno ( mysql ) = = ER_BAD_TABLE_ERROR | |
mysql_errno ( mysql ) = = ER_BAD_DB_ERROR | |
mysql_errno ( mysql ) = = ER_UNKNOWN_TABLE )
{
fprintf ( md_result_file ,
" \n -- \n -- Not dumping tablespaces as no INFORMATION_SCHEMA.FILES "
" table on this server \n -- \n " ) ;
check_io ( md_result_file ) ;
2007-08-02 06:49:29 +02:00
DBUG_RETURN ( 0 ) ;
2006-10-16 16:15:58 +02:00
}
2007-08-02 06:49:29 +02:00
my_printf_error ( 0 , " Error: '%s' when trying to dump tablespaces " ,
2006-10-16 16:15:58 +02:00
MYF ( 0 ) , mysql_error ( mysql ) ) ;
2007-08-02 06:49:29 +02:00
DBUG_RETURN ( 1 ) ;
2006-10-16 16:15:58 +02:00
}
2006-02-02 00:52:58 +01:00
buf [ 0 ] = 0 ;
while ( ( row = mysql_fetch_row ( tableres ) ) )
{
if ( strcmp ( buf , row [ 0 ] ) ! = 0 )
first = 1 ;
if ( first )
{
2012-01-10 09:03:45 +01:00
print_comment ( md_result_file , 0 , " \n -- \n -- Logfile group: %s \n -- \n " ,
row [ 0 ] ) ;
2006-02-02 00:52:58 +01:00
fprintf ( md_result_file , " \n CREATE " ) ;
}
else
{
fprintf ( md_result_file , " \n ALTER " ) ;
}
fprintf ( md_result_file ,
" LOGFILE GROUP %s \n "
2006-02-02 10:56:04 +01:00
" ADD UNDOFILE '%s' \n " ,
row [ 0 ] ,
row [ 1 ] ) ;
if ( first )
{
2006-09-19 09:07:21 +02:00
ubs = strstr ( row [ 5 ] , extra_format ) ;
if ( ! ubs )
break ;
ubs + = strlen ( extra_format ) ;
endsemi = strstr ( ubs , " ; " ) ;
if ( endsemi )
endsemi [ 0 ] = ' \0 ' ;
2006-02-02 10:56:04 +01:00
fprintf ( md_result_file ,
" UNDO_BUFFER_SIZE %s \n " ,
2006-09-19 09:07:21 +02:00
ubs ) ;
2006-02-02 10:56:04 +01:00
}
fprintf ( md_result_file ,
2006-02-02 00:52:58 +01:00
" INITIAL_SIZE %s \n "
" ENGINE=%s; \n " ,
2006-02-02 10:56:04 +01:00
row [ 3 ] ,
row [ 4 ] ) ;
2006-02-02 00:52:58 +01:00
check_io ( md_result_file ) ;
2006-02-02 10:56:04 +01:00
if ( first )
{
first = 0 ;
strxmov ( buf , row [ 0 ] , NullS ) ;
}
2006-02-02 00:52:58 +01:00
}
2006-10-16 17:02:12 +02:00
dynstr_free ( & sqlbuf ) ;
2007-08-02 06:49:29 +02:00
mysql_free_result ( tableres ) ;
2007-03-23 01:36:58 +01:00
init_dynamic_string_checked ( & sqlbuf ,
2006-10-16 17:02:12 +02:00
" SELECT DISTINCT TABLESPACE_NAME, "
" FILE_NAME, "
" LOGFILE_GROUP_NAME, "
" EXTENT_SIZE, "
" INITIAL_SIZE, "
" ENGINE "
" FROM INFORMATION_SCHEMA.FILES "
" WHERE FILE_TYPE = 'DATAFILE' " ,
256 , 1024 ) ;
if ( ts_where )
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & sqlbuf , ts_where ) ;
2006-10-16 17:02:12 +02:00
2007-03-23 01:36:58 +01:00
dynstr_append_checked ( & sqlbuf , " ORDER BY TABLESPACE_NAME, LOGFILE_GROUP_NAME " ) ;
2006-10-16 17:02:12 +02:00
if ( mysql_query_with_error_report ( mysql , & tableres , sqlbuf . str ) )
2007-08-02 06:49:29 +02:00
{
dynstr_free ( & sqlbuf ) ;
DBUG_RETURN ( 1 ) ;
}
2006-02-02 00:52:58 +01:00
buf [ 0 ] = 0 ;
while ( ( row = mysql_fetch_row ( tableres ) ) )
{
if ( strcmp ( buf , row [ 0 ] ) ! = 0 )
first = 1 ;
if ( first )
{
2012-01-10 09:03:45 +01:00
print_comment ( md_result_file , 0 , " \n -- \n -- Tablespace: %s \n -- \n " , row [ 0 ] ) ;
2006-02-02 00:52:58 +01:00
fprintf ( md_result_file , " \n CREATE " ) ;
}
else
{
fprintf ( md_result_file , " \n ALTER " ) ;
}
fprintf ( md_result_file ,
" TABLESPACE %s \n "
" ADD DATAFILE '%s' \n " ,
row [ 0 ] ,
2006-02-02 10:56:04 +01:00
row [ 1 ] ) ;
2006-02-02 00:52:58 +01:00
if ( first )
{
fprintf ( md_result_file ,
2006-02-02 10:56:04 +01:00
" USE LOGFILE GROUP %s \n "
" EXTENT_SIZE %s \n " ,
row [ 2 ] ,
row [ 3 ] ) ;
2006-02-02 00:52:58 +01:00
}
fprintf ( md_result_file ,
" INITIAL_SIZE %s \n "
" ENGINE=%s; \n " ,
2006-02-02 10:56:04 +01:00
row [ 4 ] ,
row [ 5 ] ) ;
2006-02-02 00:52:58 +01:00
check_io ( md_result_file ) ;
2006-02-02 10:56:04 +01:00
if ( first )
{
first = 0 ;
strxmov ( buf , row [ 0 ] , NullS ) ;
}
2006-02-02 00:52:58 +01:00
}
2006-10-16 17:02:12 +02:00
2007-08-02 06:49:29 +02:00
mysql_free_result ( tableres ) ;
2006-10-16 17:02:12 +02:00
dynstr_free ( & sqlbuf ) ;
2007-08-02 06:49:29 +02:00
DBUG_RETURN ( 0 ) ;
2006-02-02 00:52:58 +01:00
}
2000-07-31 21:29:14 +02:00
static int dump_all_databases ( )
{
MYSQL_ROW row ;
MYSQL_RES * tableres ;
int result = 0 ;
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & tableres , " SHOW DATABASES " ) )
2000-07-31 21:29:14 +02:00
return 1 ;
2006-08-02 03:18:47 +02:00
while ( ( row = mysql_fetch_row ( tableres ) ) )
2000-07-31 21:29:14 +02:00
{
2009-12-10 04:19:51 +01:00
if ( mysql_get_server_version ( mysql ) > = FIRST_INFORMATION_SCHEMA_VERSION & &
! my_strcasecmp ( & my_charset_latin1 , row [ 0 ] , INFORMATION_SCHEMA_DB_NAME ) )
continue ;
if ( mysql_get_server_version ( mysql ) > = FIRST_PERFORMANCE_SCHEMA_VERSION & &
! my_strcasecmp ( & my_charset_latin1 , row [ 0 ] , PERFORMANCE_SCHEMA_DB_NAME ) )
2009-05-12 19:38:14 +02:00
continue ;
2000-07-31 21:29:14 +02:00
if ( dump_all_tables_in_db ( row [ 0 ] ) )
result = 1 ;
}
2006-05-25 10:00:28 +02:00
if ( seen_views )
2004-11-12 04:01:46 +01:00
{
2006-07-24 12:00:44 +02:00
if ( mysql_query ( mysql , " SHOW DATABASES " ) | |
2006-09-11 17:14:40 +02:00
! ( tableres = mysql_store_result ( mysql ) ) )
2004-11-12 04:01:46 +01:00
{
my_printf_error ( 0 , " Error: Couldn't execute 'SHOW DATABASES': %s " ,
2006-07-24 12:00:44 +02:00
MYF ( 0 ) , mysql_error ( mysql ) ) ;
2004-11-12 04:01:46 +01:00
return 1 ;
}
2006-08-02 03:18:47 +02:00
while ( ( row = mysql_fetch_row ( tableres ) ) )
2004-11-12 04:01:46 +01:00
{
2009-12-10 04:19:51 +01:00
if ( mysql_get_server_version ( mysql ) > = FIRST_INFORMATION_SCHEMA_VERSION & &
! my_strcasecmp ( & my_charset_latin1 , row [ 0 ] , INFORMATION_SCHEMA_DB_NAME ) )
continue ;
if ( mysql_get_server_version ( mysql ) > = FIRST_PERFORMANCE_SCHEMA_VERSION & &
! my_strcasecmp ( & my_charset_latin1 , row [ 0 ] , PERFORMANCE_SCHEMA_DB_NAME ) )
2009-05-12 19:38:14 +02:00
continue ;
2004-11-12 04:01:46 +01:00
if ( dump_all_views_in_db ( row [ 0 ] ) )
result = 1 ;
}
}
2000-07-31 21:29:14 +02:00
return result ;
}
/* dump_all_databases */
static int dump_databases ( char * * db_names )
{
int result = 0 ;
2004-11-12 04:01:46 +01:00
char * * db ;
2006-11-20 21:42:06 +01:00
DBUG_ENTER ( " dump_databases " ) ;
2004-11-12 04:01:46 +01:00
for ( db = db_names ; * db ; db + + )
2001-12-20 19:14:55 +01:00
{
2004-11-12 04:01:46 +01:00
if ( dump_all_tables_in_db ( * db ) )
2000-07-31 21:29:14 +02:00
result = 1 ;
}
2006-05-25 10:00:28 +02:00
if ( ! result & & seen_views )
2004-11-12 04:01:46 +01:00
{
for ( db = db_names ; * db ; db + + )
{
if ( dump_all_views_in_db ( * db ) )
result = 1 ;
}
}
2006-11-20 21:42:06 +01:00
DBUG_RETURN ( result ) ;
2000-07-31 21:29:14 +02:00
} /* dump_databases */
2006-08-28 23:48:06 +02:00
/*
View Specific database initalization .
SYNOPSIS
init_dumping_views
qdatabase quoted name of the database
RETURN VALUES
0 Success .
1 Failure .
*/
2006-11-20 21:42:06 +01:00
int init_dumping_views ( char * qdatabase __attribute__ ( ( unused ) ) )
2006-08-28 23:48:06 +02:00
{
return 0 ;
} /* init_dumping_views */
/*
Table Specific database initalization .
SYNOPSIS
init_dumping_tables
qdatabase quoted name of the database
RETURN VALUES
0 Success .
1 Failure .
*/
2007-08-02 06:49:29 +02:00
2006-08-28 23:48:06 +02:00
int init_dumping_tables ( char * qdatabase )
{
2007-08-02 06:49:29 +02:00
DBUG_ENTER ( " init_dumping_tables " ) ;
2006-08-28 23:48:06 +02:00
if ( ! opt_create_db )
{
char qbuf [ 256 ] ;
MYSQL_ROW row ;
MYSQL_RES * dbinfo ;
my_snprintf ( qbuf , sizeof ( qbuf ) ,
" SHOW CREATE DATABASE IF NOT EXISTS %s " ,
qdatabase ) ;
if ( mysql_query ( mysql , qbuf ) | | ! ( dbinfo = mysql_store_result ( mysql ) ) )
{
/* Old server version, dump generic CREATE DATABASE */
if ( opt_drop_database )
fprintf ( md_result_file ,
2007-08-31 13:59:07 +02:00
" \n /*!40000 DROP DATABASE IF EXISTS %s*/; \n " ,
2006-08-28 23:48:06 +02:00
qdatabase ) ;
fprintf ( md_result_file ,
" \n CREATE DATABASE /*!32312 IF NOT EXISTS*/ %s; \n " ,
qdatabase ) ;
}
else
{
if ( opt_drop_database )
fprintf ( md_result_file ,
" \n /*!40000 DROP DATABASE IF EXISTS %s*/; \n " ,
qdatabase ) ;
row = mysql_fetch_row ( dbinfo ) ;
if ( row [ 1 ] )
{
fprintf ( md_result_file , " \n %s; \n " , row [ 1 ] ) ;
}
2007-08-02 06:49:29 +02:00
mysql_free_result ( dbinfo ) ;
2006-08-28 23:48:06 +02:00
}
}
2007-08-02 06:49:29 +02:00
DBUG_RETURN ( 0 ) ;
2006-08-28 23:48:06 +02:00
} /* init_dumping_tables */
static int init_dumping ( char * database , int init_func ( char * ) )
2000-07-31 21:29:14 +02:00
{
2006-07-24 12:00:44 +02:00
if ( mysql_select_db ( mysql , database ) )
2000-07-31 21:29:14 +02:00
{
2006-07-24 12:00:44 +02:00
DB_error ( mysql , " when selecting the database " ) ;
2006-05-25 10:00:28 +02:00
return 1 ; /* If --force */
2000-07-31 21:29:14 +02:00
}
2002-11-18 17:50:42 +01:00
if ( ! path & & ! opt_xml )
2000-07-31 21:29:14 +02:00
{
if ( opt_databases | | opt_alldbs )
{
2003-11-04 08:40:36 +01:00
/*
2006-05-25 10:00:28 +02:00
length of table name * 2 ( if name contains quotes ) , 2 quotes and 0
2003-11-04 08:40:36 +01:00
*/
2006-07-18 00:07:08 +02:00
char quoted_database_buf [ NAME_LEN * 2 + 3 ] ;
2003-10-24 23:26:26 +02:00
char * qdatabase = quote_name ( database , quoted_database_buf , opt_quoted ) ;
2012-01-10 09:03:45 +01:00
print_comment ( md_result_file , 0 ,
" \n -- \n -- Current Database: %s \n -- \n " , qdatabase ) ;
2003-10-14 12:19:54 +02:00
2006-08-28 23:48:06 +02:00
/* Call the view or table specific function */
init_func ( qdatabase ) ;
2003-10-14 12:19:54 +02:00
2003-11-04 08:40:36 +01:00
fprintf ( md_result_file , " \n USE %s; \n " , qdatabase ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
2000-07-31 21:29:14 +02:00
}
}
2007-03-22 18:35:29 +01:00
if ( extended_insert )
init_dynamic_string_checked ( & extended_row , " " , 1024 , 1024 ) ;
2000-07-31 21:29:14 +02:00
return 0 ;
} /* init_dumping */
2006-11-20 21:42:06 +01:00
/* Return 1 if we should copy the table */
2007-08-13 23:22:34 +02:00
my_bool include_table ( const uchar * hash_key , size_t len )
2004-12-27 19:10:30 +01:00
{
2009-10-14 18:37:38 +02:00
return ! my_hash_search ( & ignore_table , hash_key , len ) ;
2004-12-27 19:10:30 +01:00
}
2003-11-04 08:40:36 +01:00
2000-07-31 21:29:14 +02:00
static int dump_all_tables_in_db ( char * database )
{
char * table ;
uint numrows ;
2003-09-03 17:48:10 +02:00
char table_buff [ NAME_LEN * 2 + 3 ] ;
2004-12-27 19:10:30 +01:00
char hash_key [ 2 * NAME_LEN + 2 ] ; /* "db.tablename" */
char * afterdot ;
2013-05-17 15:24:36 +02:00
my_bool general_log_table_exists = 0 , slow_log_table_exists = 0 ;
int using_mysql_db = ! my_strcasecmp ( charset_info , database , " mysql " ) ;
2006-11-20 21:42:06 +01:00
DBUG_ENTER ( " dump_all_tables_in_db " ) ;
2004-12-27 19:10:30 +01:00
afterdot = strmov ( hash_key , database ) ;
* afterdot + + = ' . ' ;
2006-08-28 23:48:06 +02:00
if ( init_dumping ( database , init_dumping_tables ) )
2006-11-27 00:47:38 +01:00
DBUG_RETURN ( 1 ) ;
2002-11-18 17:50:42 +01:00
if ( opt_xml )
2006-10-19 00:43:51 +02:00
print_xml_tag ( md_result_file , " " , " \n " , " database " , " name= " , database , NullS ) ;
2009-05-12 19:38:14 +02:00
2000-07-31 21:29:14 +02:00
if ( lock_tables )
{
DYNAMIC_STRING query ;
2007-03-22 18:35:29 +01:00
init_dynamic_string_checked ( & query , " LOCK TABLES " , 256 , 1024 ) ;
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
for ( numrows = 0 ; ( table = getTableName ( 1 ) ) ; )
2000-07-31 21:29:14 +02:00
{
2007-04-05 21:53:02 +02:00
char * end = strmov ( afterdot , table ) ;
2007-08-13 15:11:25 +02:00
if ( include_table ( ( uchar * ) hash_key , end - hash_key ) )
2007-04-05 21:53:02 +02:00
{
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
numrows + + ;
2007-04-07 21:03:35 +02:00
dynstr_append_checked ( & query , quote_name ( table , table_buff , 1 ) ) ;
dynstr_append_checked ( & query , " READ /*!32311 LOCAL */, " ) ;
2007-04-05 21:53:02 +02:00
}
2000-07-31 21:29:14 +02:00
}
2006-07-24 12:00:44 +02:00
if ( numrows & & mysql_real_query ( mysql , query . str , query . length - 1 ) )
DB_error ( mysql , " when using LOCK TABLES " ) ;
2000-07-31 21:29:14 +02:00
/* We shall continue here, if --force was given */
dynstr_free ( & query ) ;
}
if ( flush_logs )
{
2006-07-24 12:00:44 +02:00
if ( mysql_refresh ( mysql , REFRESH_LOG ) )
DB_error ( mysql , " when doing refresh " ) ;
2000-07-31 21:29:14 +02:00
/* We shall continue here, if --force was given */
2011-12-23 18:35:00 +01:00
else
verbose_msg ( " -- dump_all_tables_in_db : logs flushed successfully! \n " ) ;
2000-07-31 21:29:14 +02:00
}
2004-12-27 19:10:30 +01:00
while ( ( table = getTableName ( 0 ) ) )
2000-07-31 21:29:14 +02:00
{
2004-12-27 19:10:30 +01:00
char * end = strmov ( afterdot , table ) ;
2007-08-13 15:11:25 +02:00
if ( include_table ( ( uchar * ) hash_key , end - hash_key ) )
2004-12-27 19:10:30 +01:00
{
2005-09-27 07:43:09 +02:00
dump_table ( table , database ) ;
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( order_by ) ;
2004-12-27 19:10:30 +01:00
order_by = 0 ;
2012-01-10 09:03:45 +01:00
if ( opt_dump_triggers & & mysql_get_server_version ( mysql ) > = 50009 )
2007-08-03 09:50:17 +02:00
{
if ( dump_triggers_for_table ( table , database ) )
{
if ( path )
my_fclose ( md_result_file , MYF ( MY_WME ) ) ;
maybe_exit ( EX_MYSQLERR ) ;
}
}
2004-12-27 19:10:30 +01:00
}
2013-05-17 15:24:36 +02:00
else
{
/*
If general_log and slow_log exists in the ' mysql ' database ,
we should dump the table structure . But we cannot
call get_table_structure ( ) here as ' LOCK TABLES ' query got executed
above on the session and that ' LOCK TABLES ' query does not contain
' general_log ' and ' slow_log ' tables . ( you cannot acquire lock
on log tables ) . Hence mark the existence of these log tables here and
after ' UNLOCK TABLES ' query is executed on the session , get the table
structure from server and dump it in the file .
*/
if ( using_mysql_db )
{
if ( ! my_strcasecmp ( charset_info , table , " general_log " ) )
general_log_table_exists = 1 ;
else if ( ! my_strcasecmp ( charset_info , table , " slow_log " ) )
slow_log_table_exists = 1 ;
}
}
2000-07-31 21:29:14 +02:00
}
2012-01-10 09:03:45 +01:00
if ( opt_events & & mysql_get_server_version ( mysql ) > = 50106 )
2006-03-09 21:12:43 +01:00
{
DBUG_PRINT ( " info " , ( " Dumping events for database %s " , database ) ) ;
dump_events_for_db ( database ) ;
}
2012-01-10 09:03:45 +01:00
if ( opt_routines & & mysql_get_server_version ( mysql ) > = 50009 )
2005-09-04 01:34:58 +02:00
{
DBUG_PRINT ( " info " , ( " Dumping routines for database %s " , database ) ) ;
dump_routines_for_db ( database ) ;
}
2002-11-18 17:50:42 +01:00
if ( opt_xml )
2004-06-24 22:14:12 +02:00
{
2003-10-30 11:58:30 +01:00
fputs ( " </database> \n " , md_result_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
}
2000-07-31 21:29:14 +02:00
if ( lock_tables )
2009-11-24 14:54:59 +01:00
( void ) mysql_query_with_error_report ( mysql , 0 , " UNLOCK TABLES " ) ;
2013-05-17 15:24:36 +02:00
if ( using_mysql_db )
{
char table_type [ NAME_LEN ] ;
char ignore_flag ;
if ( general_log_table_exists )
{
if ( ! get_table_structure ( ( char * ) " general_log " ,
database , table_type , & ignore_flag ) )
verbose_msg ( " -- Warning: get_table_structure() failed with some internal "
" error for 'general_log' table \n " ) ;
}
if ( slow_log_table_exists )
{
if ( ! get_table_structure ( ( char * ) " slow_log " ,
database , table_type , & ignore_flag ) )
verbose_msg ( " -- Warning: get_table_structure() failed with some internal "
" error for 'slow_log' table \n " ) ;
}
}
if ( flush_privileges & & using_mysql_db )
2006-09-14 20:56:14 +02:00
{
fprintf ( md_result_file , " \n -- \n -- Flush Grant Tables \n -- \n " ) ;
fprintf ( md_result_file , " \n /*! FLUSH PRIVILEGES */; \n " ) ;
}
2006-11-27 00:47:38 +01:00
DBUG_RETURN ( 0 ) ;
2000-07-31 21:29:14 +02:00
} /* dump_all_tables_in_db */
2004-11-12 04:01:46 +01:00
/*
dump structure of views of database
SYNOPSIS
dump_all_views_in_db ( )
database database name
2000-07-31 21:29:14 +02:00
2004-11-12 04:01:46 +01:00
RETURN
0 OK
1 ERROR
*/
static my_bool dump_all_views_in_db ( char * database )
{
char * table ;
uint numrows ;
char table_buff [ NAME_LEN * 2 + 3 ] ;
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
char hash_key [ 2 * NAME_LEN + 2 ] ; /* "db.tablename" */
char * afterdot ;
afterdot = strmov ( hash_key , database ) ;
* afterdot + + = ' . ' ;
2004-11-12 04:01:46 +01:00
2006-08-28 23:48:06 +02:00
if ( init_dumping ( database , init_dumping_views ) )
2006-07-14 01:25:13 +02:00
return 1 ;
2004-11-12 04:01:46 +01:00
if ( opt_xml )
2006-10-19 00:43:51 +02:00
print_xml_tag ( md_result_file , " " , " \n " , " database " , " name= " , database , NullS ) ;
2004-11-12 04:01:46 +01:00
if ( lock_tables )
{
DYNAMIC_STRING query ;
2007-03-22 18:35:29 +01:00
init_dynamic_string_checked ( & query , " LOCK TABLES " , 256 , 1024 ) ;
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
for ( numrows = 0 ; ( table = getTableName ( 1 ) ) ; )
2004-11-12 04:01:46 +01:00
{
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
char * end = strmov ( afterdot , table ) ;
2009-07-07 15:11:46 +02:00
if ( include_table ( ( uchar * ) hash_key , end - hash_key ) )
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
{
numrows + + ;
dynstr_append_checked ( & query , quote_name ( table , table_buff , 1 ) ) ;
dynstr_append_checked ( & query , " READ /*!32311 LOCAL */, " ) ;
}
2004-11-12 04:01:46 +01:00
}
2006-07-24 12:00:44 +02:00
if ( numrows & & mysql_real_query ( mysql , query . str , query . length - 1 ) )
DB_error ( mysql , " when using LOCK TABLES " ) ;
2004-11-12 04:01:46 +01:00
/* We shall continue here, if --force was given */
dynstr_free ( & query ) ;
}
if ( flush_logs )
{
2006-07-24 12:00:44 +02:00
if ( mysql_refresh ( mysql , REFRESH_LOG ) )
DB_error ( mysql , " when doing refresh " ) ;
2004-11-12 04:01:46 +01:00
/* We shall continue here, if --force was given */
2011-12-23 18:35:00 +01:00
else
verbose_msg ( " -- dump_all_views_in_db : logs flushed successfully! \n " ) ;
2004-11-12 04:01:46 +01:00
}
while ( ( table = getTableName ( 0 ) ) )
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
{
char * end = strmov ( afterdot , table ) ;
2009-07-07 15:11:46 +02:00
if ( include_table ( ( uchar * ) hash_key , end - hash_key ) )
WL#3984 (Revise locking of mysql.general_log and mysql.slow_log)
Bug#25422 (Hang with log tables)
Bug 17876 (Truncating mysql.slow_log in a SP after using cursor locks the
thread)
Bug 23044 (Warnings on flush of a log table)
Bug 29129 (Resetting general_log while the GLOBAL READ LOCK is set causes
a deadlock)
Prior to this fix, the server would hang when performing concurrent
ALTER TABLE or TRUNCATE TABLE statements against the LOG TABLES,
which are mysql.general_log and mysql.slow_log.
The root cause traces to the following code:
in sql_base.cc, open_table()
if (table->in_use != thd)
{
/* wait_for_condition will unlock LOCK_open for us */
wait_for_condition(thd, &LOCK_open, &COND_refresh);
}
The problem with this code is that the current implementation of the
LOGGER creates 'fake' THD objects, like
- Log_to_csv_event_handler::general_log_thd
- Log_to_csv_event_handler::slow_log_thd
which are not associated to a real thread running in the server,
so that waiting for these non-existing threads to release table locks
cause the dead lock.
In general, the design of Log_to_csv_event_handler does not fit into the
general architecture of the server, so that the concept of general_log_thd
and slow_log_thd has to be abandoned:
- this implementation does not work with table locking
- it will not work with commands like SHOW PROCESSLIST
- having the log tables always opened does not integrate well with DDL
operations / FLUSH TABLES / SET GLOBAL READ_ONLY
With this patch, the fundamental design of the LOGGER has been changed to:
- always open and close a log table when writing a log
- remove totally the usage of fake THD objects
- clarify how locking of log tables is implemented in general.
See WL#3984 for details related to the new locking design.
Additional changes (misc bugs exposed and fixed):
1)
mysqldump which would ignore some tables in dump_all_tables_in_db(),
but forget to ignore the same in dump_all_views_in_db().
2)
mysqldump would also issue an empty "LOCK TABLE" command when all the tables
to lock are to be ignored (numrows == 0), instead of not issuing the query.
3)
Internal errors handlers could intercept errors but not warnings
(see sql_error.cc).
4)
Implementing a nested call to open tables, for the performance schema tables,
exposed an existing bug in remove_table_from_cache(), which would perform:
in_use->some_tables_deleted=1;
against another thread, without any consideration about thread locking.
This call inside remove_table_from_cache() was not required anyway,
since calling mysql_lock_abort() takes care of aborting -- cleanly -- threads
that might hold a lock on a table.
This line (in_use->some_tables_deleted=1) has been removed.
2007-07-27 08:31:06 +02:00
get_view_structure ( table , database ) ;
}
2004-11-12 04:01:46 +01:00
if ( opt_xml )
{
fputs ( " </database> \n " , md_result_file ) ;
check_io ( md_result_file ) ;
}
if ( lock_tables )
2009-11-24 14:54:59 +01:00
( void ) mysql_query_with_error_report ( mysql , 0 , " UNLOCK TABLES " ) ;
2004-11-12 04:01:46 +01:00
return 0 ;
} /* dump_all_tables_in_db */
2000-07-31 21:29:14 +02:00
2005-03-17 14:51:07 +01:00
2005-01-27 21:48:26 +01:00
/*
2006-01-03 17:54:54 +01:00
get_actual_table_name - - executes a SHOW TABLES LIKE ' % s ' to get the actual
table name from the server for the table name given on the command line .
we do this because the table name given on the command line may be a
2005-01-29 16:25:56 +01:00
different case ( e . g . T1 vs t1 )
2006-01-03 17:54:54 +01:00
2005-01-27 21:48:26 +01:00
RETURN
2006-05-29 08:17:38 +02:00
pointer to the table name
0 if error
2005-01-27 21:48:26 +01:00
*/
2006-05-29 08:17:38 +02:00
static char * get_actual_table_name ( const char * old_table_name , MEM_ROOT * root )
2005-02-01 15:27:08 +01:00
{
2006-05-29 08:17:38 +02:00
char * name = 0 ;
2005-04-04 23:32:48 +02:00
MYSQL_RES * table_res ;
2005-02-01 15:27:08 +01:00
MYSQL_ROW row ;
2005-03-16 21:56:56 +01:00
char query [ 50 + 2 * NAME_LEN ] ;
2005-03-16 13:57:59 +01:00
char show_name_buff [ FN_REFLEN ] ;
2005-02-01 15:27:08 +01:00
DBUG_ENTER ( " get_actual_table_name " ) ;
2005-01-27 21:48:26 +01:00
2005-03-16 21:56:56 +01:00
/* Check memory for quote_for_like() */
DBUG_ASSERT ( 2 * sizeof ( old_table_name ) < sizeof ( show_name_buff ) ) ;
2006-01-03 17:54:54 +01:00
my_snprintf ( query , sizeof ( query ) , " SHOW TABLES LIKE %s " ,
2006-05-25 10:00:28 +02:00
quote_for_like ( old_table_name , show_name_buff ) ) ;
2005-03-16 21:56:56 +01:00
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , 0 , query ) )
2007-03-22 18:35:29 +01:00
return NullS ;
2005-01-27 21:48:26 +01:00
2006-07-24 12:00:44 +02:00
if ( ( table_res = mysql_store_result ( mysql ) ) )
2005-03-28 21:06:43 +02:00
{
2005-04-04 23:32:48 +02:00
my_ulonglong num_rows = mysql_num_rows ( table_res ) ;
if ( num_rows > 0 )
{
2006-05-29 09:32:08 +02:00
ulong * lengths ;
2005-04-04 23:32:48 +02:00
/*
Return first row
TODO : Return all matching rows
*/
row = mysql_fetch_row ( table_res ) ;
2006-05-29 09:32:08 +02:00
lengths = mysql_fetch_lengths ( table_res ) ;
2006-05-29 08:17:38 +02:00
name = strmake_root ( root , row [ 0 ] , lengths [ 0 ] ) ;
2005-04-04 23:32:48 +02:00
}
mysql_free_result ( table_res ) ;
2005-03-28 21:06:43 +02:00
}
2006-05-29 08:17:38 +02:00
DBUG_PRINT ( " exit " , ( " new_table_name: %s " , name ) ) ;
DBUG_RETURN ( name ) ;
2005-02-01 15:27:08 +01:00
}
2005-01-27 21:48:26 +01:00
2000-07-31 21:29:14 +02:00
static int dump_selected_tables ( char * db , char * * table_names , int tables )
{
2007-04-03 13:13:27 +02:00
char table_buff [ NAME_LEN * 2 + 3 ] ;
2005-06-21 14:19:56 +02:00
DYNAMIC_STRING lock_tables_query ;
2006-05-29 08:17:38 +02:00
MEM_ROOT root ;
char * * dump_tables , * * pos , * * end ;
2005-06-21 14:19:56 +02:00
DBUG_ENTER ( " dump_selected_tables " ) ;
2000-07-31 21:29:14 +02:00
2006-08-28 23:48:06 +02:00
if ( init_dumping ( db , init_dumping_tables ) )
2006-02-16 12:31:33 +01:00
DBUG_RETURN ( 1 ) ;
2005-06-21 14:19:56 +02:00
2006-05-29 08:17:38 +02:00
init_alloc_root ( & root , 8192 , 0 ) ;
if ( ! ( dump_tables = pos = ( char * * ) alloc_root ( & root , tables * sizeof ( char * ) ) ) )
2007-03-22 18:35:29 +01:00
die ( EX_EOM , " alloc_root failure. " ) ;
2005-06-21 14:19:56 +02:00
2007-03-22 18:35:29 +01:00
init_dynamic_string_checked ( & lock_tables_query , " LOCK TABLES " , 256 , 1024 ) ;
2005-06-21 14:19:56 +02:00
for ( ; tables > 0 ; tables - - , table_names + + )
2000-07-31 21:29:14 +02:00
{
2005-06-21 14:19:56 +02:00
/* the table name passed on commandline may be wrong case */
2006-05-29 08:17:38 +02:00
if ( ( * pos = get_actual_table_name ( * table_names , & root ) ) )
2000-07-31 21:29:14 +02:00
{
2005-06-21 14:19:56 +02:00
/* Add found table name to lock_tables_query */
if ( lock_tables )
{
2007-03-22 18:35:29 +01:00
dynstr_append_checked ( & lock_tables_query , quote_name ( * pos , table_buff , 1 ) ) ;
dynstr_append_checked ( & lock_tables_query , " READ /*!32311 LOCAL */, " ) ;
2005-06-21 14:19:56 +02:00
}
2006-05-29 08:17:38 +02:00
pos + + ;
2000-07-31 21:29:14 +02:00
}
2005-06-21 14:19:56 +02:00
else
2000-07-31 21:29:14 +02:00
{
2007-08-02 06:49:29 +02:00
if ( ! ignore_errors )
{
dynstr_free ( & lock_tables_query ) ;
free_root ( & root , MYF ( 0 ) ) ;
}
maybe_die ( EX_ILLEGAL_TABLE , " Couldn't find table: \" %s \" " , * table_names ) ;
/* We shall countinue here, if --force was given */
2000-07-31 21:29:14 +02:00
}
2005-06-21 14:19:56 +02:00
}
2006-05-29 08:17:38 +02:00
end = pos ;
2005-06-21 14:19:56 +02:00
2009-12-10 04:19:51 +01:00
/* Can't LOCK TABLES in I_S / P_S, so don't try. */
2009-05-12 19:38:14 +02:00
if ( lock_tables & &
2009-12-10 04:19:51 +01:00
! ( mysql_get_server_version ( mysql ) > = FIRST_INFORMATION_SCHEMA_VERSION & &
! my_strcasecmp ( & my_charset_latin1 , db , INFORMATION_SCHEMA_DB_NAME ) ) & &
! ( mysql_get_server_version ( mysql ) > = FIRST_PERFORMANCE_SCHEMA_VERSION & &
! my_strcasecmp ( & my_charset_latin1 , db , PERFORMANCE_SCHEMA_DB_NAME ) ) )
2005-06-21 14:19:56 +02:00
{
2006-07-24 12:00:44 +02:00
if ( mysql_real_query ( mysql , lock_tables_query . str ,
2005-06-21 14:19:56 +02:00
lock_tables_query . length - 1 ) )
2007-08-02 06:49:29 +02:00
{
if ( ! ignore_errors )
{
dynstr_free ( & lock_tables_query ) ;
free_root ( & root , MYF ( 0 ) ) ;
}
2006-07-24 12:00:44 +02:00
DB_error ( mysql , " when doing LOCK TABLES " ) ;
2000-07-31 21:29:14 +02:00
/* We shall countinue here, if --force was given */
2007-08-02 06:49:29 +02:00
}
2000-07-31 21:29:14 +02:00
}
2005-06-21 14:19:56 +02:00
dynstr_free ( & lock_tables_query ) ;
2000-07-31 21:29:14 +02:00
if ( flush_logs )
{
2006-07-24 12:00:44 +02:00
if ( mysql_refresh ( mysql , REFRESH_LOG ) )
2007-08-02 06:49:29 +02:00
{
if ( ! ignore_errors )
free_root ( & root , MYF ( 0 ) ) ;
2006-07-24 12:00:44 +02:00
DB_error ( mysql , " when doing refresh " ) ;
2007-08-02 06:49:29 +02:00
}
2000-07-31 21:29:14 +02:00
/* We shall countinue here, if --force was given */
2011-12-23 18:35:00 +01:00
else
verbose_msg ( " -- dump_selected_tables : logs flushed successfully! \n " ) ;
2000-07-31 21:29:14 +02:00
}
2002-11-18 17:50:42 +01:00
if ( opt_xml )
2006-10-19 00:43:51 +02:00
print_xml_tag ( md_result_file , " " , " \n " , " database " , " name= " , db , NullS ) ;
2005-06-22 15:43:49 +02:00
2005-06-21 14:19:56 +02:00
/* Dump each selected table */
2006-05-29 08:17:38 +02:00
for ( pos = dump_tables ; pos < end ; pos + + )
2000-07-31 21:29:14 +02:00
{
2006-05-29 08:17:38 +02:00
DBUG_PRINT ( " info " , ( " Dumping table %s " , * pos ) ) ;
dump_table ( * pos , db ) ;
2005-09-09 00:03:07 +02:00
if ( opt_dump_triggers & &
2006-07-24 12:00:44 +02:00
mysql_get_server_version ( mysql ) > = 50009 )
2007-08-03 09:50:17 +02:00
{
if ( dump_triggers_for_table ( * pos , db ) )
{
if ( path )
my_fclose ( md_result_file , MYF ( MY_WME ) ) ;
maybe_exit ( EX_MYSQLERR ) ;
}
}
2004-11-12 04:01:46 +01:00
}
2005-06-22 15:43:49 +02:00
/* Dump each selected view */
2006-05-25 10:00:28 +02:00
if ( seen_views )
2004-11-12 04:01:46 +01:00
{
2006-05-29 08:17:38 +02:00
for ( pos = dump_tables ; pos < end ; pos + + )
get_view_structure ( * pos , db ) ;
2000-07-31 21:29:14 +02:00
}
2012-01-10 09:03:45 +01:00
if ( opt_events & & mysql_get_server_version ( mysql ) > = 50106 )
2006-03-09 21:12:43 +01:00
{
DBUG_PRINT ( " info " , ( " Dumping events for database %s " , db ) ) ;
dump_events_for_db ( db ) ;
}
2005-09-04 01:34:58 +02:00
/* obtain dump of routines (procs/functions) */
2012-01-10 09:03:45 +01:00
if ( opt_routines & & mysql_get_server_version ( mysql ) > = 50009 )
2005-09-04 01:34:58 +02:00
{
DBUG_PRINT ( " info " , ( " Dumping routines for database %s " , db ) ) ;
dump_routines_for_db ( db ) ;
}
2006-05-29 08:17:38 +02:00
free_root ( & root , MYF ( 0 ) ) ;
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( order_by ) ;
2005-06-21 14:19:56 +02:00
order_by = 0 ;
2002-11-18 17:50:42 +01:00
if ( opt_xml )
2004-06-24 22:14:12 +02:00
{
2003-10-30 11:58:30 +01:00
fputs ( " </database> \n " , md_result_file ) ;
2004-06-24 22:14:12 +02:00
check_io ( md_result_file ) ;
}
2000-07-31 21:29:14 +02:00
if ( lock_tables )
2009-11-24 14:54:59 +01:00
( void ) mysql_query_with_error_report ( mysql , 0 , " UNLOCK TABLES " ) ;
2005-06-21 14:19:56 +02:00
DBUG_RETURN ( 0 ) ;
2000-07-31 21:29:14 +02:00
} /* dump_selected_tables */
2004-11-10 17:56:45 +01:00
static int do_show_master_status ( MYSQL * mysql_con )
{
MYSQL_ROW row ;
MYSQL_RES * master ;
const char * comment_prefix =
( opt_master_data = = MYSQL_OPT_MASTER_DATA_COMMENTED_SQL ) ? " -- " : " " ;
if ( mysql_query_with_error_report ( mysql_con , & master , " SHOW MASTER STATUS " ) )
{
return 1 ;
}
else
{
2006-08-02 03:18:47 +02:00
row = mysql_fetch_row ( master ) ;
2004-11-10 17:56:45 +01:00
if ( row & & row [ 0 ] & & row [ 1 ] )
{
2005-07-01 12:01:00 +02:00
/* SHOW MASTER STATUS reports file and position */
2012-01-10 09:03:45 +01:00
print_comment ( md_result_file , 0 ,
" \n -- \n -- Position to start replication or point-in-time "
" recovery from \n -- \n \n " ) ;
2004-11-10 17:56:45 +01:00
fprintf ( md_result_file ,
" %sCHANGE MASTER TO MASTER_LOG_FILE='%s', MASTER_LOG_POS=%s; \n " ,
2006-05-25 10:00:28 +02:00
comment_prefix , row [ 0 ] , row [ 1 ] ) ;
2004-11-10 17:56:45 +01:00
check_io ( md_result_file ) ;
}
2005-07-01 12:01:00 +02:00
else if ( ! ignore_errors )
{
/* SHOW MASTER STATUS reports nothing and --force is not enabled */
2006-05-25 10:00:28 +02:00
my_printf_error ( 0 , " Error: Binlogging on server not active " ,
MYF ( 0 ) ) ;
2005-07-01 12:01:00 +02:00
mysql_free_result ( master ) ;
2008-03-07 10:15:49 +01:00
maybe_exit ( EX_MYSQLERR ) ;
2005-07-01 12:01:00 +02:00
return 1 ;
}
2004-11-10 17:56:45 +01:00
mysql_free_result ( master ) ;
}
return 0 ;
}
2009-11-04 14:31:03 +01:00
static int do_stop_slave_sql ( MYSQL * mysql_con )
{
MYSQL_RES * slave ;
/* We need to check if the slave sql is running in the first place */
if ( mysql_query_with_error_report ( mysql_con , & slave , " SHOW SLAVE STATUS " ) )
return ( 1 ) ;
else
{
MYSQL_ROW row = mysql_fetch_row ( slave ) ;
if ( row & & row [ 11 ] )
{
/* if SLAVE SQL is not running, we don't stop it */
if ( ! strcmp ( row [ 11 ] , " No " ) )
{
mysql_free_result ( slave ) ;
/* Silently assume that they don't have the slave running */
return ( 0 ) ;
}
}
}
mysql_free_result ( slave ) ;
/* now, stop slave if running */
if ( mysql_query_with_error_report ( mysql_con , 0 , " STOP SLAVE SQL_THREAD " ) )
return ( 1 ) ;
return ( 0 ) ;
}
static int add_stop_slave ( void )
{
if ( opt_comments )
fprintf ( md_result_file ,
" \n -- \n -- stop slave statement to make a recovery dump) \n -- \n \n " ) ;
fprintf ( md_result_file , " STOP SLAVE; \n " ) ;
return ( 0 ) ;
}
static int add_slave_statements ( void )
{
if ( opt_comments )
fprintf ( md_result_file ,
" \n -- \n -- start slave statement to make a recovery dump) \n -- \n \n " ) ;
fprintf ( md_result_file , " START SLAVE; \n " ) ;
return ( 0 ) ;
}
static int do_show_slave_status ( MYSQL * mysql_con )
{
MYSQL_RES * slave ;
const char * comment_prefix =
( opt_slave_data = = MYSQL_OPT_SLAVE_DATA_COMMENTED_SQL ) ? " -- " : " " ;
if ( mysql_query_with_error_report ( mysql_con , & slave , " SHOW SLAVE STATUS " ) )
{
if ( ! ignore_errors )
{
/* SHOW SLAVE STATUS reports nothing and --force is not enabled */
my_printf_error ( 0 , " Error: Slave not set up " , MYF ( 0 ) ) ;
}
mysql_free_result ( slave ) ;
return 1 ;
}
else
{
MYSQL_ROW row = mysql_fetch_row ( slave ) ;
if ( row & & row [ 9 ] & & row [ 21 ] )
{
/* SHOW MASTER STATUS reports file and position */
if ( opt_comments )
fprintf ( md_result_file ,
" \n -- \n -- Position to start replication or point-in-time "
" recovery from (the master of this slave) \n -- \n \n " ) ;
fprintf ( md_result_file , " %sCHANGE MASTER TO " , comment_prefix ) ;
if ( opt_include_master_host_port )
{
if ( row [ 1 ] )
fprintf ( md_result_file , " MASTER_HOST='%s', " , row [ 1 ] ) ;
if ( row [ 3 ] )
2013-04-12 10:48:21 +02:00
fprintf ( md_result_file , " MASTER_PORT=%s, " , row [ 3 ] ) ;
2009-11-04 14:31:03 +01:00
}
fprintf ( md_result_file ,
" MASTER_LOG_FILE='%s', MASTER_LOG_POS=%s; \n " , row [ 9 ] , row [ 21 ] ) ;
check_io ( md_result_file ) ;
}
mysql_free_result ( slave ) ;
}
return 0 ;
}
static int do_start_slave_sql ( MYSQL * mysql_con )
{
MYSQL_RES * slave ;
/* We need to check if the slave sql is stopped in the first place */
if ( mysql_query_with_error_report ( mysql_con , & slave , " SHOW SLAVE STATUS " ) )
return ( 1 ) ;
else
{
MYSQL_ROW row = mysql_fetch_row ( slave ) ;
if ( row & & row [ 11 ] )
{
/* if SLAVE SQL is not running, we don't start it */
if ( ! strcmp ( row [ 11 ] , " Yes " ) )
{
mysql_free_result ( slave ) ;
/* Silently assume that they don't have the slave running */
return ( 0 ) ;
}
}
}
mysql_free_result ( slave ) ;
/* now, start slave if stopped */
if ( mysql_query_with_error_report ( mysql_con , 0 , " START SLAVE " ) )
{
my_printf_error ( 0 , " Error: Unable to start slave " , MYF ( 0 ) ) ;
return 1 ;
}
return ( 0 ) ;
}
2004-11-10 17:56:45 +01:00
static int do_flush_tables_read_lock ( MYSQL * mysql_con )
{
2004-11-14 17:45:37 +01:00
/*
We do first a FLUSH TABLES . If a long update is running , the FLUSH TABLES
will wait but will not stall the whole mysqld , and when the long update is
done the FLUSH TABLES WITH READ LOCK will start and succeed quickly . So ,
FLUSH TABLES is to lower the probability of a stage where both mysqldump
and most client connections are stalled . Of course , if a second long
update starts between the two FLUSHes , we have that bad stall .
*/
2006-05-25 10:00:28 +02:00
return
2008-04-14 23:43:08 +02:00
( mysql_query_with_error_report ( mysql_con , 0 ,
( ( opt_master_data ! = 0 ) ?
" FLUSH /*!40101 LOCAL */ TABLES " :
" FLUSH TABLES " ) ) | |
2004-11-14 17:45:37 +01:00
mysql_query_with_error_report ( mysql_con , 0 ,
" FLUSH TABLES WITH READ LOCK " ) ) ;
2004-11-10 17:56:45 +01:00
}
static int do_unlock_tables ( MYSQL * mysql_con )
{
return mysql_query_with_error_report ( mysql_con , 0 , " UNLOCK TABLES " ) ;
}
2007-05-31 00:37:41 +02:00
static int get_bin_log_name ( MYSQL * mysql_con ,
char * buff_log_name , uint buff_len )
2007-05-27 17:50:10 +02:00
{
MYSQL_RES * res ;
MYSQL_ROW row ;
2007-05-31 00:37:41 +02:00
if ( mysql_query ( mysql_con , " SHOW MASTER STATUS " ) | |
2007-05-27 17:50:10 +02:00
! ( res = mysql_store_result ( mysql ) ) )
return 1 ;
if ( ! ( row = mysql_fetch_row ( res ) ) )
{
mysql_free_result ( res ) ;
return 1 ;
}
/*
Only one row is returned , and the first column is the name of the
active log .
*/
strmake ( buff_log_name , row [ 0 ] , buff_len - 1 ) ;
2004-11-10 17:56:45 +01:00
2007-05-27 17:50:10 +02:00
mysql_free_result ( res ) ;
return 0 ;
}
static int purge_bin_logs_to ( MYSQL * mysql_con , char * log_name )
2004-11-10 17:56:45 +01:00
{
2007-05-27 17:50:10 +02:00
DYNAMIC_STRING str ;
int err ;
init_dynamic_string_checked ( & str , " PURGE BINARY LOGS TO ' " , 1024 , 1024 ) ;
dynstr_append_checked ( & str , log_name ) ;
dynstr_append_checked ( & str , " ' " ) ;
err = mysql_query_with_error_report ( mysql_con , 0 , str . str ) ;
dynstr_free ( & str ) ;
return err ;
2004-11-10 17:56:45 +01:00
}
2006-10-03 17:26:50 +02:00
static int start_transaction ( MYSQL * mysql_con )
2004-11-10 17:56:45 +01:00
{
2011-12-23 18:35:00 +01:00
verbose_msg ( " -- Starting transaction... \n " ) ;
2004-11-10 17:56:45 +01:00
/*
We use BEGIN for old servers . - - single - transaction - - master - data will fail
on old servers , but that ' s ok as it was already silently broken ( it didn ' t
do a consistent read , so better tell people frankly , with the error ) .
2005-01-13 00:44:13 +01:00
We want the first consistent read to be used for all tables to dump so we
need the REPEATABLE READ level ( not anything lower , for example READ
COMMITTED would give one new consistent read per dumped table ) .
2004-11-10 17:56:45 +01:00
*/
2007-10-04 08:27:03 +02:00
if ( ( mysql_get_server_version ( mysql_con ) < 40100 ) & & opt_master_data )
{
fprintf ( stderr , " -- %s: the combination of --single-transaction and "
" --master-data requires a MySQL server version of at least 4.1 "
" (current server's version is %s). %s \n " ,
ignore_errors ? " Warning " : " Error " ,
mysql_con - > server_version ? mysql_con - > server_version : " unknown " ,
ignore_errors ? " Continuing due to --force, backup may not be consistent across all tables! " : " Aborting. " ) ;
if ( ! ignore_errors )
exit ( EX_MYSQLERR ) ;
}
2004-11-10 17:56:45 +01:00
return ( mysql_query_with_error_report ( mysql_con , 0 ,
2005-01-13 00:44:13 +01:00
" SET SESSION TRANSACTION ISOLATION "
" LEVEL REPEATABLE READ " ) | |
mysql_query_with_error_report ( mysql_con , 0 ,
2004-11-10 17:56:45 +01:00
" START TRANSACTION "
2006-10-03 17:26:50 +02:00
" /*!40100 WITH CONSISTENT SNAPSHOT */ " ) ) ;
2004-11-10 17:56:45 +01:00
}
2003-01-16 04:35:59 +01:00
static ulong find_set ( TYPELIB * lib , const char * x , uint length ,
2006-05-25 10:00:28 +02:00
char * * err_pos , uint * err_len )
2003-01-16 04:35:59 +01:00
{
const char * end = x + length ;
ulong found = 0 ;
uint find ;
char buff [ 255 ] ;
2003-03-06 13:51:37 +01:00
* err_pos = 0 ; /* No error yet */
2003-03-16 08:20:45 +01:00
while ( end > x & & my_isspace ( charset_info , end [ - 1 ] ) )
2003-01-16 04:35:59 +01:00
end - - ;
* err_len = 0 ;
if ( x ! = end )
{
const char * start = x ;
for ( ; ; )
{
const char * pos = start ;
uint var_len ;
for ( ; pos ! = end & & * pos ! = ' , ' ; pos + + ) ;
var_len = ( uint ) ( pos - start ) ;
2009-12-17 18:58:38 +01:00
strmake ( buff , start , min ( sizeof ( buff ) - 1 , var_len ) ) ;
2011-02-11 15:00:09 +01:00
find = find_type ( buff , lib , FIND_TYPE_BASIC ) ;
2003-01-16 04:35:59 +01:00
if ( ! find )
{
* err_pos = ( char * ) start ;
* err_len = var_len ;
}
else
found | = ( ( longlong ) 1 < < ( find - 1 ) ) ;
if ( pos = = end )
break ;
start = pos + 1 ;
}
}
return found ;
}
2000-07-31 21:29:14 +02:00
/* Print a value with a prefix on file */
static void print_value ( FILE * file , MYSQL_RES * result , MYSQL_ROW row ,
2006-05-25 10:00:28 +02:00
const char * prefix , const char * name ,
int string_value )
2000-07-31 21:29:14 +02:00
{
2006-05-25 10:00:28 +02:00
MYSQL_FIELD * field ;
2000-07-31 21:29:14 +02:00
mysql_field_seek ( result , 0 ) ;
2006-08-02 03:18:47 +02:00
for ( ; ( field = mysql_fetch_field ( result ) ) ; row + + )
2000-07-31 21:29:14 +02:00
{
if ( ! strcmp ( field - > name , name ) )
{
if ( row [ 0 ] & & row [ 0 ] [ 0 ] & & strcmp ( row [ 0 ] , " 0 " ) ) /* Skip default */
{
2006-05-25 10:00:28 +02:00
fputc ( ' ' , file ) ;
fputs ( prefix , file ) ;
if ( string_value )
unescape ( file , row [ 0 ] , ( uint ) strlen ( row [ 0 ] ) ) ;
else
fputs ( row [ 0 ] , file ) ;
check_io ( file ) ;
return ;
2000-07-31 21:29:14 +02:00
}
}
}
2006-05-25 10:00:28 +02:00
return ; /* This shouldn't happen */
2000-07-31 21:29:14 +02:00
} /* print_value */
2003-12-14 05:39:52 +01:00
/*
SYNOPSIS
2005-09-27 07:43:09 +02:00
Check if we the table is one of the table types that should be ignored :
MRG_ISAM , MRG_MYISAM , if opt_delayed , if that table supports delayed inserts .
If the table should be altogether ignored , it returns a TRUE , FALSE if it
2006-05-25 10:00:28 +02:00
should not be ignored . If the user has selected to use INSERT DELAYED , it
sets the value of the bool pointer supports_delayed_inserts to 0 if not
2005-09-27 07:43:09 +02:00
supported , 1 if it is supported .
ARGS
2003-12-14 05:39:52 +01:00
check_if_ignore_table ( )
2006-05-25 10:00:28 +02:00
table_name Table name to check
2005-09-27 07:43:09 +02:00
table_type Type of table
2003-12-14 05:39:52 +01:00
GLOBAL VARIABLES
2006-07-24 12:00:44 +02:00
mysql MySQL connection
2006-05-25 10:00:28 +02:00
verbose Write warning messages
2003-12-14 05:39:52 +01:00
RETURN
2005-09-27 07:43:09 +02:00
char ( bit value ) See IGNORE_ values at top
2003-12-14 05:39:52 +01:00
*/
2005-09-27 07:43:09 +02:00
char check_if_ignore_table ( const char * table_name , char * table_type )
2003-12-14 05:39:52 +01:00
{
2005-09-27 18:40:51 +02:00
char result = IGNORE_NONE ;
2003-12-14 05:39:52 +01:00
char buff [ FN_REFLEN + 80 ] , show_name_buff [ FN_REFLEN ] ;
2007-03-23 23:15:11 +01:00
MYSQL_RES * res = NULL ;
2003-12-14 05:39:52 +01:00
MYSQL_ROW row ;
2005-09-27 07:43:09 +02:00
DBUG_ENTER ( " check_if_ignore_table " ) ;
2003-12-14 05:39:52 +01:00
2005-03-16 21:56:56 +01:00
/* Check memory for quote_for_like() */
DBUG_ASSERT ( 2 * sizeof ( table_name ) < sizeof ( show_name_buff ) ) ;
my_snprintf ( buff , sizeof ( buff ) , " show table status like %s " ,
2006-05-25 10:00:28 +02:00
quote_for_like ( table_name , show_name_buff ) ) ;
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & res , buff ) )
2003-12-14 05:39:52 +01:00
{
2006-07-24 12:00:44 +02:00
if ( mysql_errno ( mysql ) ! = ER_PARSE_ERROR )
2006-05-25 10:00:28 +02:00
{ /* If old MySQL version */
2006-11-20 21:42:06 +01:00
verbose_msg ( " -- Warning: Couldn't get status information for "
" table %s (%s) \n " , table_name , mysql_error ( mysql ) ) ;
2005-09-27 18:40:51 +02:00
DBUG_RETURN ( result ) ; /* assume table is ok */
2003-12-14 05:39:52 +01:00
}
}
2004-11-10 17:56:45 +01:00
if ( ! ( row = mysql_fetch_row ( res ) ) )
2003-12-14 05:39:52 +01:00
{
fprintf ( stderr ,
2006-05-25 10:00:28 +02:00
" Error: Couldn't read status information for table %s (%s) \n " ,
2006-07-24 12:00:44 +02:00
table_name , mysql_error ( mysql ) ) ;
2005-01-24 15:48:25 +01:00
mysql_free_result ( res ) ;
2005-09-27 18:40:51 +02:00
DBUG_RETURN ( result ) ; /* assume table is ok */
2003-12-14 05:39:52 +01:00
}
2004-11-12 04:01:46 +01:00
if ( ! ( row [ 1 ] ) )
2005-10-11 23:58:22 +02:00
strmake ( table_type , " VIEW " , NAME_LEN - 1 ) ;
2004-11-12 04:01:46 +01:00
else
{
2005-09-27 07:43:09 +02:00
/*
If the table type matches any of these , we do support delayed inserts .
Note : we do not want to skip dumping this table if if is not one of
these types , but we do want to use delayed inserts in the dump if
the table type is _NOT_ one of these types
*/
2005-09-27 18:40:51 +02:00
strmake ( table_type , row [ 1 ] , NAME_LEN - 1 ) ;
2005-09-27 07:43:09 +02:00
if ( opt_delayed )
{
if ( strcmp ( table_type , " MyISAM " ) & &
strcmp ( table_type , " ISAM " ) & &
strcmp ( table_type , " ARCHIVE " ) & &
strcmp ( table_type , " HEAP " ) & &
strcmp ( table_type , " MEMORY " ) )
result = IGNORE_INSERT_DELAYED ;
}
/*
If these two types , we do want to skip dumping the table
*/
2006-07-24 11:56:01 +02:00
if ( ! opt_no_data & &
2007-08-03 00:14:05 +02:00
( ! my_strcasecmp ( & my_charset_latin1 , table_type , " MRG_MyISAM " ) | |
2009-03-19 09:49:51 +01:00
! strcmp ( table_type , " MRG_ISAM " ) | |
2009-03-19 07:06:37 +01:00
! strcmp ( table_type , " FEDERATED " ) ) )
2005-09-27 07:43:09 +02:00
result = IGNORE_DATA ;
2004-11-12 04:01:46 +01:00
}
2005-01-24 15:48:25 +01:00
mysql_free_result ( res ) ;
2005-09-27 07:43:09 +02:00
DBUG_RETURN ( result ) ;
2003-12-14 05:39:52 +01:00
}
2006-07-24 12:00:44 +02:00
2004-11-16 23:45:24 +01:00
/*
Get string of comma - separated primary key field names
SYNOPSIS
char * primary_key_fields ( const char * table_name )
RETURNS pointer to allocated buffer ( must be freed by caller )
table_name quoted table name
DESCRIPTION
Use SHOW KEYS FROM table_name , allocate a buffer to hold the
field names , and then build that string and return the pointer
to that buffer .
Returns NULL if there is no PRIMARY or UNIQUE key on the table ,
or if there is some failure . It is better to continue to dump
the table unsorted , rather than exit without dumping the data .
*/
2005-07-18 14:33:18 +02:00
2004-11-16 23:45:24 +01:00
static char * primary_key_fields ( const char * table_name )
{
2006-08-02 03:18:47 +02:00
MYSQL_RES * res = NULL ;
2004-11-16 23:45:24 +01:00
MYSQL_ROW row ;
/* SHOW KEYS FROM + table name * 2 (escaped) + 2 quotes + \0 */
2006-07-18 00:07:08 +02:00
char show_keys_buff [ 15 + NAME_LEN * 2 + 3 ] ;
2006-08-02 03:18:47 +02:00
uint result_length = 0 ;
char * result = 0 ;
2006-08-02 03:56:08 +02:00
char buff [ NAME_LEN * 2 + 3 ] ;
char * quoted_field ;
2004-11-16 23:45:24 +01:00
2006-05-25 10:00:28 +02:00
my_snprintf ( show_keys_buff , sizeof ( show_keys_buff ) ,
" SHOW KEYS FROM %s " , table_name ) ;
2006-07-24 12:00:44 +02:00
if ( mysql_query ( mysql , show_keys_buff ) | |
2006-09-11 17:14:40 +02:00
! ( res = mysql_store_result ( mysql ) ) )
2004-11-16 23:45:24 +01:00
{
fprintf ( stderr , " Warning: Couldn't read keys from table %s; "
" records are NOT sorted (%s) \n " ,
2006-07-24 12:00:44 +02:00
table_name , mysql_error ( mysql ) ) ;
2004-11-16 23:45:24 +01:00
/* Don't exit, because it's better to print out unsorted records */
goto cleanup ;
}
2004-11-30 02:26:24 +01:00
/*
* Figure out the length of the ORDER BY clause result .
* Note that SHOW KEYS is ordered : a PRIMARY key is always the first
* row , and UNIQUE keys come before others . So we only need to check
* the first key , not all keys .
*/
2006-08-02 03:18:47 +02:00
if ( ( row = mysql_fetch_row ( res ) ) & & atoi ( row [ 1 ] ) = = 0 )
2004-11-16 23:45:24 +01:00
{
2004-11-30 02:26:24 +01:00
/* Key is unique */
do
2006-08-02 03:56:08 +02:00
{
quoted_field = quote_name ( row [ 4 ] , buff , 0 ) ;
result_length + = strlen ( quoted_field ) + 1 ; /* + 1 for ',' or \0 */
} while ( ( row = mysql_fetch_row ( res ) ) & & atoi ( row [ 3 ] ) > 1 ) ;
2004-11-16 23:45:24 +01:00
}
/* Build the ORDER BY clause result */
2005-07-18 14:33:18 +02:00
if ( result_length )
{
2004-11-16 23:45:24 +01:00
char * end ;
/* result (terminating \0 is already in result_length) */
2006-08-02 03:18:47 +02:00
result = my_malloc ( result_length + 10 , MYF ( MY_WME ) ) ;
2005-07-18 14:33:18 +02:00
if ( ! result )
{
2004-11-16 23:45:24 +01:00
fprintf ( stderr , " Error: Not enough memory to store ORDER BY clause \n " ) ;
goto cleanup ;
}
2004-11-30 02:26:24 +01:00
mysql_data_seek ( res , 0 ) ;
2006-08-02 03:18:47 +02:00
row = mysql_fetch_row ( res ) ;
2006-08-02 03:56:08 +02:00
quoted_field = quote_name ( row [ 4 ] , buff , 0 ) ;
end = strmov ( result , quoted_field ) ;
2006-08-02 03:18:47 +02:00
while ( ( row = mysql_fetch_row ( res ) ) & & atoi ( row [ 3 ] ) > 1 )
2006-08-02 03:56:08 +02:00
{
quoted_field = quote_name ( row [ 4 ] , buff , 0 ) ;
end = strxmov ( end , " , " , quoted_field , NullS ) ;
}
2004-11-16 23:45:24 +01:00
}
cleanup :
if ( res )
mysql_free_result ( res ) ;
return result ;
}
2003-12-14 05:39:52 +01:00
2006-02-21 13:21:17 +01:00
/*
Replace a substring
SYNOPSIS
replace
ds_str The string to search and perform the replace in
search_str The string to search for
search_len Length of the string to search for
replace_str The string to replace with
replace_len Length of the string to replace with
RETURN
0 String replaced
1 Could not find search_str in str
*/
static int replace ( DYNAMIC_STRING * ds_str ,
const char * search_str , ulong search_len ,
const char * replace_str , ulong replace_len )
{
2006-02-23 14:06:29 +01:00
DYNAMIC_STRING ds_tmp ;
2006-02-21 13:21:17 +01:00
const char * start = strstr ( ds_str - > str , search_str ) ;
if ( ! start )
return 1 ;
2007-03-22 18:35:29 +01:00
init_dynamic_string_checked ( & ds_tmp , " " ,
2006-05-25 10:00:28 +02:00
ds_str - > length + replace_len , 256 ) ;
2007-03-22 18:35:29 +01:00
dynstr_append_mem_checked ( & ds_tmp , ds_str - > str , start - ds_str - > str ) ;
dynstr_append_mem_checked ( & ds_tmp , replace_str , replace_len ) ;
dynstr_append_checked ( & ds_tmp , start + search_len ) ;
dynstr_set_checked ( ds_str , ds_tmp . str ) ;
2006-02-21 13:21:17 +01:00
dynstr_free ( & ds_tmp ) ;
return 0 ;
}
2004-11-12 04:01:46 +01:00
/*
Getting VIEW structure
SYNOPSIS
2005-04-04 23:32:48 +02:00
get_view_structure ( )
2004-11-12 04:01:46 +01:00
table view name
db db name
RETURN
0 OK
1 ERROR
*/
2005-04-04 23:32:48 +02:00
static my_bool get_view_structure ( char * table , char * db )
2004-11-12 04:01:46 +01:00
{
2005-04-04 23:32:48 +02:00
MYSQL_RES * table_res ;
2004-11-12 04:01:46 +01:00
MYSQL_ROW row ;
MYSQL_FIELD * field ;
2006-05-25 10:00:28 +02:00
char * result_table , * opt_quoted_table ;
char table_buff [ NAME_LEN * 2 + 3 ] ;
char table_buff2 [ NAME_LEN * 2 + 3 ] ;
2006-02-21 13:21:17 +01:00
char query [ QUERY_LENGTH ] ;
2006-08-02 03:18:47 +02:00
FILE * sql_file = md_result_file ;
2005-04-04 23:32:48 +02:00
DBUG_ENTER ( " get_view_structure " ) ;
2004-11-12 04:01:46 +01:00
2006-07-24 11:56:01 +02:00
if ( opt_no_create_info ) /* Don't write table creation info */
2004-11-12 04:01:46 +01:00
DBUG_RETURN ( 0 ) ;
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- Retrieving view structure for table %s... \n " , table ) ;
2004-11-12 04:01:46 +01:00
2005-05-19 00:40:34 +02:00
# ifdef NOT_REALLY_USED_YET
2013-07-18 08:10:08 +02:00
sprintf ( insert_pat , " SET SQL_QUOTE_SHOW_CREATE=%d " ,
2006-05-25 10:00:28 +02:00
( opt_quoted | | opt_keywords ) ) ;
2005-05-19 00:40:34 +02:00
# endif
2004-11-12 04:01:46 +01:00
result_table = quote_name ( table , table_buff , 1 ) ;
opt_quoted_table = quote_name ( table , table_buff2 , 0 ) ;
2007-07-12 10:49:39 +02:00
if ( switch_character_set_results ( mysql , " binary " ) )
DBUG_RETURN ( 1 ) ;
2006-02-23 16:08:28 +01:00
my_snprintf ( query , sizeof ( query ) , " SHOW CREATE TABLE %s " , result_table ) ;
2007-07-12 10:49:39 +02:00
2006-07-24 12:00:44 +02:00
if ( mysql_query_with_error_report ( mysql , & table_res , query ) )
2007-07-12 10:49:39 +02:00
{
switch_character_set_results ( mysql , default_charset ) ;
2004-11-12 04:01:46 +01:00
DBUG_RETURN ( 0 ) ;
2007-07-12 10:49:39 +02:00
}
2004-11-12 04:01:46 +01:00
2006-02-21 13:21:17 +01:00
/* Check if this is a view */
field = mysql_fetch_field_direct ( table_res , 0 ) ;
if ( strcmp ( field - > name , " View " ) ! = 0 )
{
2007-07-12 10:49:39 +02:00
switch_character_set_results ( mysql , default_charset ) ;
2006-07-24 11:51:52 +02:00
verbose_msg ( " -- It's base table, skipped \n " ) ;
2006-02-21 13:21:17 +01:00
DBUG_RETURN ( 0 ) ;
}
/* If requested, open separate .sql file for this view */
2004-11-12 04:01:46 +01:00
if ( path )
{
2009-05-27 15:16:21 +02:00
if ( ! ( sql_file = open_sql_file_for_table ( table , O_WRONLY ) ) )
2004-11-12 04:01:46 +01:00
DBUG_RETURN ( 1 ) ;
2007-03-22 18:35:29 +01:00
2004-11-12 04:01:46 +01:00
write_header ( sql_file , db ) ;
}
2012-01-10 09:03:45 +01:00
print_comment ( sql_file , 0 ,
" \n -- \n -- Final view structure for view %s \n -- \n \n " ,
result_table ) ;
2009-05-18 21:52:51 +02:00
/* Table might not exist if this view was dumped with --tab. */
fprintf ( sql_file , " /*!50001 DROP TABLE IF EXISTS %s*/; \n " , opt_quoted_table ) ;
2004-11-12 04:01:46 +01:00
if ( opt_drop )
{
2005-07-29 22:39:08 +02:00
fprintf ( sql_file , " /*!50001 DROP VIEW IF EXISTS %s*/; \n " ,
opt_quoted_table ) ;
2004-11-12 04:01:46 +01:00
check_io ( sql_file ) ;
}
2006-02-23 16:08:28 +01:00
my_snprintf ( query , sizeof ( query ) ,
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
" SELECT CHECK_OPTION, DEFINER, SECURITY_TYPE, "
" CHARACTER_SET_CLIENT, COLLATION_CONNECTION "
" FROM information_schema.views "
2006-02-23 16:08:28 +01:00
" WHERE table_name= \" %s \" AND table_schema= \" %s \" " , table , db ) ;
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
2006-07-24 12:00:44 +02:00
if ( mysql_query ( mysql , query ) )
2006-02-21 13:21:17 +01:00
{
/*
Use the raw output from SHOW CREATE TABLE if
information_schema query fails .
*/
row = mysql_fetch_row ( table_res ) ;
fprintf ( sql_file , " /*!50001 %s */; \n " , row [ 1 ] ) ;
check_io ( sql_file ) ;
mysql_free_result ( table_res ) ;
}
else
{
char * ptr ;
ulong * lengths ;
char search_buf [ 256 ] , replace_buf [ 256 ] ;
ulong search_len , replace_len ;
DYNAMIC_STRING ds_view ;
/* Save the result of SHOW CREATE TABLE in ds_view */
row = mysql_fetch_row ( table_res ) ;
lengths = mysql_fetch_lengths ( table_res ) ;
2007-03-22 18:35:29 +01:00
init_dynamic_string_checked ( & ds_view , row [ 1 ] , lengths [ 1 ] + 1 , 1024 ) ;
2006-02-21 13:21:17 +01:00
mysql_free_result ( table_res ) ;
/* Get the result from "select ... information_schema" */
2006-07-24 12:00:44 +02:00
if ( ! ( table_res = mysql_store_result ( mysql ) ) | |
2006-07-14 01:25:13 +02:00
! ( row = mysql_fetch_row ( table_res ) ) )
2006-02-21 13:21:17 +01:00
{
2007-08-02 06:49:29 +02:00
if ( table_res )
mysql_free_result ( table_res ) ;
dynstr_free ( & ds_view ) ;
2007-03-22 18:35:29 +01:00
DB_error ( mysql , " when trying to save the result of SHOW CREATE TABLE in ds_view. " ) ;
2006-02-21 13:21:17 +01:00
DBUG_RETURN ( 1 ) ;
}
2006-07-14 01:25:13 +02:00
2006-02-21 13:21:17 +01:00
lengths = mysql_fetch_lengths ( table_res ) ;
/*
" WITH %s CHECK OPTION " is available from 5.0 .2
Surround it with ! 50002 comments
*/
if ( strcmp ( row [ 0 ] , " NONE " ) )
{
ptr = search_buf ;
search_len = ( ulong ) ( strxmov ( ptr , " WITH " , row [ 0 ] ,
" CHECK OPTION " , NullS ) - ptr ) ;
ptr = replace_buf ;
replace_len = ( ulong ) ( strxmov ( ptr , " */ \n /*!50002 WITH " , row [ 0 ] ,
" CHECK OPTION " , NullS ) - ptr ) ;
replace ( & ds_view , search_buf , search_len , replace_buf , replace_len ) ;
}
/*
" DEFINER=%s SQL SECURITY %s " is available from 5.0 .13
Surround it with ! 50013 comments
*/
{
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
size_t user_name_len ;
2006-02-21 13:21:17 +01:00
char user_name_str [ USERNAME_LENGTH + 1 ] ;
char quoted_user_name_str [ USERNAME_LENGTH * 2 + 3 ] ;
WL#3817: Simplify string / memory area types and make things more consistent (first part)
The following type conversions was done:
- Changed byte to uchar
- Changed gptr to uchar*
- Change my_string to char *
- Change my_size_t to size_t
- Change size_s to size_t
Removed declaration of byte, gptr, my_string, my_size_t and size_s.
Following function parameter changes was done:
- All string functions in mysys/strings was changed to use size_t
instead of uint for string lengths.
- All read()/write() functions changed to use size_t (including vio).
- All protocoll functions changed to use size_t instead of uint
- Functions that used a pointer to a string length was changed to use size_t*
- Changed malloc(), free() and related functions from using gptr to use void *
as this requires fewer casts in the code and is more in line with how the
standard functions work.
- Added extra length argument to dirname_part() to return the length of the
created string.
- Changed (at least) following functions to take uchar* as argument:
- db_dump()
- my_net_write()
- net_write_command()
- net_store_data()
- DBUG_DUMP()
- decimal2bin() & bin2decimal()
- Changed my_compress() and my_uncompress() to use size_t. Changed one
argument to my_uncompress() from a pointer to a value as we only return
one value (makes function easier to use).
- Changed type of 'pack_data' argument to packfrm() to avoid casts.
- Changed in readfrm() and writefrom(), ha_discover and handler::discover()
the type for argument 'frmdata' to uchar** to avoid casts.
- Changed most Field functions to use uchar* instead of char* (reduced a lot of
casts).
- Changed field->val_xxx(xxx, new_ptr) to take const pointers.
Other changes:
- Removed a lot of not needed casts
- Added a few new cast required by other changes
- Added some cast to my_multi_malloc() arguments for safety (as string lengths
needs to be uint, not size_t).
- Fixed all calls to hash-get-key functions to use size_t*. (Needed to be done
explicitely as this conflict was often hided by casting the function to
hash_get_key).
- Changed some buffers to memory regions to uchar* to avoid casts.
- Changed some string lengths from uint to size_t.
- Changed field->ptr to be uchar* instead of char*. This allowed us to
get rid of a lot of casts.
- Some changes from true -> TRUE, false -> FALSE, unsigned char -> uchar
- Include zlib.h in some files as we needed declaration of crc32()
- Changed MY_FILE_ERROR to be (size_t) -1.
- Changed many variables to hold the result of my_read() / my_write() to be
size_t. This was needed to properly detect errors (which are
returned as (size_t) -1).
- Removed some very old VMS code
- Changed packfrm()/unpackfrm() to not be depending on uint size
(portability fix)
- Removed windows specific code to restore cursor position as this
causes slowdown on windows and we should not mix read() and pread()
calls anyway as this is not thread safe. Updated function comment to
reflect this. Changed function that depended on original behavior of
my_pwrite() to itself restore the cursor position (one such case).
- Added some missing checking of return value of malloc().
- Changed definition of MOD_PAD_CHAR_TO_FULL_LENGTH to avoid 'long' overflow.
- Changed type of table_def::m_size from my_size_t to ulong to reflect that
m_size is the number of elements in the array, not a string/memory
length.
- Moved THD::max_row_length() to table.cc (as it's not depending on THD).
Inlined max_row_length_blob() into this function.
- More function comments
- Fixed some compiler warnings when compiled without partitions.
- Removed setting of LEX_STRING() arguments in declaration (portability fix).
- Some trivial indentation/variable name changes.
- Some trivial code simplifications:
- Replaced some calls to alloc_root + memcpy to use
strmake_root()/strdup_root().
- Changed some calls from memdup() to strmake() (Safety fix)
- Simpler loops in client-simple.c
2007-05-10 11:59:39 +02:00
size_t host_name_len ;
2006-02-21 13:21:17 +01:00
char host_name_str [ HOSTNAME_LENGTH + 1 ] ;
char quoted_host_name_str [ HOSTNAME_LENGTH * 2 + 3 ] ;
parse_user ( row [ 1 ] , lengths [ 1 ] , user_name_str , & user_name_len ,
2006-05-25 10:00:28 +02:00
host_name_str , & host_name_len ) ;
2006-02-21 13:21:17 +01:00
ptr = search_buf ;
search_len =
( ulong ) ( strxmov ( ptr , " DEFINER= " ,
quote_name ( user_name_str , quoted_user_name_str , FALSE ) ,
" @ " ,
quote_name ( host_name_str , quoted_host_name_str , FALSE ) ,
" SQL SECURITY " , row [ 2 ] , NullS ) - ptr ) ;
ptr = replace_buf ;
replace_len =
( ulong ) ( strxmov ( ptr , " */ \n /*!50013 DEFINER= " ,
quote_name ( user_name_str , quoted_user_name_str , FALSE ) ,
" @ " ,
quote_name ( host_name_str , quoted_host_name_str , FALSE ) ,
" SQL SECURITY " , row [ 2 ] ,
" */ \n /*!50001 " , NullS ) - ptr ) ;
replace ( & ds_view , search_buf , search_len , replace_buf , replace_len ) ;
}
/* Dump view structure to file */
Patch for the following bugs:
- BUG#11986: Stored routines and triggers can fail if the code
has a non-ascii symbol
- BUG#16291: mysqldump corrupts string-constants with non-ascii-chars
- BUG#19443: INFORMATION_SCHEMA does not support charsets properly
- BUG#21249: Character set of SP-var can be ignored
- BUG#25212: Character set of string constant is ignored (stored routines)
- BUG#25221: Character set of string constant is ignored (triggers)
There were a few general problems that caused these bugs:
1. Character set information of the original (definition) query for views,
triggers, stored routines and events was lost.
2. mysqldump output query in client character set, which can be
inappropriate to encode definition-query.
3. INFORMATION_SCHEMA used strings with mixed encodings to display object
definition;
1. No query-definition-character set.
In order to compile query into execution code, some extra data (such as
environment variables or the database character set) is used. The problem
here was that this context was not preserved. So, on the next load it can
differ from the original one, thus the result will be different.
The context contains the following data:
- client character set;
- connection collation (character set and collation);
- collation of the owner database;
The fix is to store this context and use it each time we parse (compile)
and execute the object (stored routine, trigger, ...).
2. Wrong mysqldump-output.
The original query can contain several encodings (by means of character set
introducers). The problem here was that we tried to convert original query
to the mysqldump-client character set.
Moreover, we stored queries in different character sets for different
objects (views, for one, used UTF8, triggers used original character set).
The solution is
- to store definition queries in the original character set;
- to change SHOW CREATE statement to output definition query in the
binary character set (i.e. without any conversion);
- introduce SHOW CREATE TRIGGER statement;
- to dump special statements to switch the context to the original one
before dumping and restore it afterwards.
Note, in order to preserve the database collation at the creation time,
additional ALTER DATABASE might be used (to temporary switch the database
collation back to the original value). In this case, ALTER DATABASE
privilege will be required. This is a backward-incompatible change.
3. INFORMATION_SCHEMA showed non-UTF8 strings
The fix is to generate UTF8-query during the parsing, store it in the object
and show it in the INFORMATION_SCHEMA.
Basically, the idea is to create a copy of the original query convert it to
UTF8. Character set introducers are removed and all text literals are
converted to UTF8.
This UTF8 query is intended to provide user-readable output. It must not be
used to recreate the object. Specialized SHOW CREATE statements should be
used for this.
The reason for this limitation is the following: the original query can
contain symbols from several character sets (by means of character set
introducers).
Example:
- original query:
CREATE VIEW v1 AS SELECT _cp1251 'Hello' AS c1;
- UTF8 query (for INFORMATION_SCHEMA):
CREATE VIEW v1 AS SELECT 'Hello' AS c1;
2007-06-28 19:34:54 +02:00
fprintf ( sql_file ,
" /*!50001 SET @saved_cs_client = @@character_set_client */; \n "
" /*!50001 SET @saved_cs_results = @@character_set_results */; \n "
" /*!50001 SET @saved_col_connection = @@collation_connection */; \n "
" /*!50001 SET character_set_client = %s */; \n "
" /*!50001 SET character_set_results = %s */; \n "
" /*!50001 SET collation_connection = %s */; \n "
" /*!50001 %s */; \n "
" /*!50001 SET character_set_client = @saved_cs_client */; \n "
" /*!50001 SET character_set_results = @saved_cs_results */; \n "
" /*!50001 SET collation_connection = @saved_col_connection */; \n " ,
( const char * ) row [ 3 ] ,
( const char * ) row [ 3 ] ,
( const char * ) row [ 4 ] ,
( const char * ) ds_view . str ) ;
2006-02-21 13:21:17 +01:00
check_io ( sql_file ) ;
mysql_free_result ( table_res ) ;
dynstr_free ( & ds_view ) ;
}
2007-07-12 10:49:39 +02:00
if ( switch_character_set_results ( mysql , default_charset ) )
DBUG_RETURN ( 1 ) ;
2006-02-21 13:21:17 +01:00
/* If a separate .sql file was opened, close it now */
2004-11-12 04:01:46 +01:00
if ( sql_file ! = md_result_file )
{
fputs ( " \n " , sql_file ) ;
write_footer ( sql_file ) ;
my_fclose ( sql_file , MYF ( MY_WME ) ) ;
}
DBUG_RETURN ( 0 ) ;
}
2007-03-22 18:35:29 +01:00
/*
The following functions are wrappers for the dynamic string functions
and if they fail , the wrappers will terminate the current process .
*/
# define DYNAMIC_STR_ERROR_MSG "Couldn't perform DYNAMIC_STRING operation"
static void init_dynamic_string_checked ( DYNAMIC_STRING * str , const char * init_str ,
uint init_alloc , uint alloc_increment )
{
if ( init_dynamic_string ( str , init_str , init_alloc , alloc_increment ) )
die ( EX_MYSQLERR , DYNAMIC_STR_ERROR_MSG ) ;
}
static void dynstr_append_checked ( DYNAMIC_STRING * dest , const char * src )
{
if ( dynstr_append ( dest , src ) )
die ( EX_MYSQLERR , DYNAMIC_STR_ERROR_MSG ) ;
}
static void dynstr_set_checked ( DYNAMIC_STRING * str , const char * init_str )
{
if ( dynstr_set ( str , init_str ) )
die ( EX_MYSQLERR , DYNAMIC_STR_ERROR_MSG ) ;
}
static void dynstr_append_mem_checked ( DYNAMIC_STRING * str , const char * append ,
uint length )
{
if ( dynstr_append_mem ( str , append , length ) )
die ( EX_MYSQLERR , DYNAMIC_STR_ERROR_MSG ) ;
}
static void dynstr_realloc_checked ( DYNAMIC_STRING * str , ulong additional_size )
{
if ( dynstr_realloc ( str , additional_size ) )
die ( EX_MYSQLERR , DYNAMIC_STR_ERROR_MSG ) ;
}
2004-11-12 04:01:46 +01:00
2000-07-31 21:29:14 +02:00
int main ( int argc , char * * argv )
{
2007-05-27 17:50:10 +02:00
char bin_log_name [ FN_REFLEN ] ;
2006-11-20 21:42:06 +01:00
int exit_code ;
2005-07-31 11:49:55 +02:00
MY_INIT ( " mysqldump " ) ;
2004-03-11 15:46:27 +01:00
compatible_mode_normal_str [ 0 ] = 0 ;
2005-01-06 18:13:58 +01:00
default_charset = ( char * ) mysql_universal_client_charset ;
2005-06-27 19:31:00 +02:00
bzero ( ( char * ) & ignore_table , sizeof ( ignore_table ) ) ;
2001-10-04 19:43:06 +02:00
2006-11-20 21:42:06 +01:00
exit_code = get_options ( & argc , & argv ) ;
if ( exit_code )
2000-07-31 21:29:14 +02:00
{
2009-08-28 17:06:59 +02:00
free_resources ( ) ;
2006-11-20 21:42:06 +01:00
exit ( exit_code ) ;
2000-07-31 21:29:14 +02:00
}
2007-03-23 19:24:03 +01:00
2012-01-10 09:03:45 +01:00
/*
Disable comments in xml mode if ' comments ' option is not explicitly used .
*/
if ( opt_xml & & ! opt_comments_used )
opt_comments = 0 ;
2007-03-23 19:24:03 +01:00
if ( log_error_file )
{
if ( ! ( stderror_file = freopen ( log_error_file , " a+ " , stderr ) ) )
{
2009-08-28 17:06:59 +02:00
free_resources ( ) ;
2007-03-23 19:24:03 +01:00
exit ( EX_MYSQLERR ) ;
}
}
2006-11-20 21:42:06 +01:00
if ( connect_to_db ( current_host , current_user , opt_password ) )
{
2009-08-28 17:06:59 +02:00
free_resources ( ) ;
2000-07-31 21:29:14 +02:00
exit ( EX_MYSQLERR ) ;
2006-11-20 21:42:06 +01:00
}
2000-07-31 21:29:14 +02:00
if ( ! path )
2002-01-02 20:29:41 +01:00
write_header ( md_result_file , * argv ) ;
2009-11-04 14:31:03 +01:00
if ( opt_slave_data & & do_stop_slave_sql ( mysql ) )
goto err ;
2011-12-23 18:35:00 +01:00
if ( ( opt_lock_all_tables | | opt_master_data | |
( opt_single_transaction & & flush_logs ) ) & &
2006-07-24 12:00:44 +02:00
do_flush_tables_read_lock ( mysql ) )
2004-11-10 17:56:45 +01:00
goto err ;
2011-12-23 18:35:00 +01:00
/*
Flush logs before starting transaction since
this causes implicit commit starting mysql - 5.5 .
*/
if ( opt_lock_all_tables | | opt_master_data | |
( opt_single_transaction & & flush_logs ) | |
opt_delete_master_logs )
2007-05-27 17:50:10 +02:00
{
2011-12-23 18:35:00 +01:00
if ( flush_logs | | opt_delete_master_logs )
{
if ( mysql_refresh ( mysql , REFRESH_LOG ) )
goto err ;
verbose_msg ( " -- main : logs flushed successfully! \n " ) ;
}
/* Not anymore! That would not be sensible. */
2007-05-27 17:50:10 +02:00
flush_logs = 0 ;
}
2011-12-23 18:35:00 +01:00
if ( opt_delete_master_logs )
2002-06-14 13:14:30 +02:00
{
2011-12-23 18:35:00 +01:00
if ( get_bin_log_name ( mysql , bin_log_name , sizeof ( bin_log_name ) ) )
2004-11-10 17:56:45 +01:00
goto err ;
2002-06-14 13:14:30 +02:00
}
2011-12-23 18:35:00 +01:00
if ( opt_single_transaction & & start_transaction ( mysql ) )
goto err ;
2009-11-04 14:31:03 +01:00
/* Add 'STOP SLAVE to beginning of dump */
if ( opt_slave_apply & & add_stop_slave ( ) )
goto err ;
2006-07-24 12:00:44 +02:00
if ( opt_master_data & & do_show_master_status ( mysql ) )
2004-11-10 17:56:45 +01:00
goto err ;
2009-11-04 14:31:03 +01:00
if ( opt_slave_data & & do_show_slave_status ( mysql ) )
goto err ;
2006-07-24 12:00:44 +02:00
if ( opt_single_transaction & & do_unlock_tables ( mysql ) ) /* unlock but no commit! */
2004-11-10 17:56:45 +01:00
goto err ;
2006-02-02 00:52:58 +01:00
if ( opt_alltspcs )
dump_all_tablespaces ( ) ;
2000-07-31 21:29:14 +02:00
if ( opt_alldbs )
2006-10-09 16:03:46 +02:00
{
if ( ! opt_alltspcs & & ! opt_notspcs )
dump_all_tablespaces ( ) ;
2000-07-31 21:29:14 +02:00
dump_all_databases ( ) ;
2006-10-09 16:03:46 +02:00
}
2000-07-31 21:29:14 +02:00
else if ( argc > 1 & & ! opt_databases )
2002-06-11 10:20:31 +02:00
{
/* Only one database and selected table(s) */
2006-10-09 16:03:46 +02:00
if ( ! opt_alltspcs & & ! opt_notspcs )
dump_tablespaces_for_tables ( * argv , ( argv + 1 ) , ( argc - 1 ) ) ;
2000-07-31 21:29:14 +02:00
dump_selected_tables ( * argv , ( argv + 1 ) , ( argc - 1 ) ) ;
2002-06-11 10:20:31 +02:00
}
2000-09-07 03:55:17 +02:00
else
2002-06-11 10:20:31 +02:00
{
/* One or more databases, all tables */
2006-10-09 16:03:46 +02:00
if ( ! opt_alltspcs & & ! opt_notspcs )
dump_tablespaces_for_databases ( argv ) ;
2000-07-31 21:29:14 +02:00
dump_databases ( argv ) ;
2002-06-11 10:20:31 +02:00
}
2007-05-27 17:50:10 +02:00
2009-11-04 14:31:03 +01:00
/* if --dump-slave , start the slave sql thread */
if ( opt_slave_data & & do_start_slave_sql ( mysql ) )
goto err ;
/* add 'START SLAVE' to end of dump */
if ( opt_slave_apply & & add_slave_statements ( ) )
goto err ;
2007-05-27 17:50:10 +02:00
/* ensure dumped data flushed */
if ( md_result_file & & fflush ( md_result_file ) )
{
if ( ! first_error )
first_error = EX_MYSQLERR ;
goto err ;
}
/* everything successful, purge the old logs files */
if ( opt_delete_master_logs & & purge_bin_logs_to ( mysql , bin_log_name ) )
goto err ;
2002-11-14 20:16:30 +01:00
# ifdef HAVE_SMEM
Bug#34043: Server loops excessively in _checkchunk() when safemalloc is enabled
Essentially, the problem is that safemalloc is excruciatingly
slow as it checks all allocated blocks for overrun at each
memory management primitive, yielding a almost exponential
slowdown for the memory management functions (malloc, realloc,
free). The overrun check basically consists of verifying some
bytes of a block for certain magic keys, which catches some
simple forms of overrun. Another minor problem is violation
of aliasing rules and that its own internal list of blocks
is prone to corruption.
Another issue with safemalloc is rather the maintenance cost
as the tool has a significant impact on the server code.
Given the magnitude of memory debuggers available nowadays,
especially those that are provided with the platform malloc
implementation, maintenance of a in-house and largely obsolete
memory debugger becomes a burden that is not worth the effort
due to its slowness and lack of support for detecting more
common forms of heap corruption.
Since there are third-party tools that can provide the same
functionality at a lower or comparable performance cost, the
solution is to simply remove safemalloc. Third-party tools
can provide the same functionality at a lower or comparable
performance cost.
The removal of safemalloc also allows a simplification of the
malloc wrappers, removing quite a bit of kludge: redefinition
of my_malloc, my_free and the removal of the unused second
argument of my_free. Since free() always check whether the
supplied pointer is null, redudant checks are also removed.
Also, this patch adds unit testing for my_malloc and moves
my_realloc implementation into the same file as the other
memory allocation primitives.
2010-07-08 23:20:08 +02:00
my_free ( shared_memory_base_name ) ;
2002-11-14 20:16:30 +01:00
# endif
2004-11-10 17:56:45 +01:00
/*
No reason to explicitely COMMIT the transaction , neither to explicitely
UNLOCK TABLES : these will be automatically be done by the server when we
disconnect now . Saves some code here , some network trips , adds nothing to
server .
*/
err :
2000-07-31 21:29:14 +02:00
dbDisconnect ( current_host ) ;
2004-02-13 22:21:46 +01:00
if ( ! path )
write_footer ( md_result_file ) ;
2006-11-20 21:42:06 +01:00
free_resources ( ) ;
2007-03-23 19:24:03 +01:00
if ( stderror_file )
fclose ( stderror_file ) ;
2000-07-31 21:29:14 +02:00
return ( first_error ) ;
} /* main */