This is the non-ndb part of the patch.
The return value of mysql_bin_log.write was ignored by most callers,
which may lead to inconsistent on master and slave if the transaction
was committed while the binlog was not correctly written. If
my_error() is call in mysql_bin_log.write, this could also lead to
assertion issue if my_ok() or my_error() is called after.
This fixed the problem by let the caller to check and handle the
return value of mysql_bin_log.write. This patch only adresses the
simple cases.
------------------------------------------------------------
revno: 2572.23.1
committer: davi@mysql.com/endora.local
timestamp: Wed 2008-03-19 09:03:08 -0300
message:
Bug#17954 Threads_connected > Threads_created
The problem is that insert delayed threads are counted as connected
but not as created, leading to a Threads_connected value greater then
the Threads_created value.
The solution is to enforce the documented behavior that the
Threads_connected value shall be the number of currently
open connections and that Threads_created shall be the
number of threads created to handle connections.
------------------------------------------------------------
revno: 2476.1116.1
committer: davi@mysql.com/endora.local
timestamp: Fri 2007-12-14 10:10:19 -0200
message:
DROP TABLE under LOCK TABLES simultaneous to a FLUSH TABLES
WITH READ LOCK (global read lock) can lead to a deadlock.
The solution is to not wait for the global read lock if the
thread is holding any locked tables.
Related to bugs 23713 and 32395. This issues is being fixed
only on 6.0 because it depends on the fix for bug 25858 --
which was fixed only on 6.0.
------------------------------------------------------------
revno: 2476.784.3
committer: davi@moksha.local
timestamp: Tue 2007-10-02 21:27:31 -0300
message:
Bug#25858 Some DROP TABLE under LOCK TABLES can cause deadlocks
When a client (connection) holds a lock on a table and attempts to
drop (obtain a exclusive lock) on a second table that is already
held by a second client and the second client then attempts to
drop the table that is held by the first client, leads to a
circular wait deadlock. This scenario is very similar to trying to
drop (or rename) a table while holding read locks and are
correctly forbidden.
The solution is to allow a drop table operation to continue only
if the table being dropped is write (exclusively) locked, or if
the table is temporary, or if the client is not holding any
locks. Using this scheme prevents the creation of a circular
chain in which each client is waiting for one table that the
next client in the chain is holding.
This is incompatible change, as can be seen by number of tests
cases that needed to be fixed, but is consistent with respect to
behavior of the different scenarios in which the circular wait
might happen.
The mentioned on the bug report set of bugs fixes have not be pushed to the main trees.
Fixed with extracting commits done to 6.0-rpl tree and applying them to the main 5.1.
Notes.
1. part of changes - the mtr's specific - were packported to the main 5.0 tree for mtr v1
as http://lists.mysql.com/commits/46562
However, there is no that fix anymore in the mtr v2. (This fact was mailed to mtr maintaining
people).
2. Bug@36929 crash in kill_zombie_dump_threads-> THD::awake() with replication tests
is not backported because the base code of the patch is libevent and that was removed
from the main trees due to its instability.
Bug #48370 Absolutely wrong calculations with GROUP BY and
decimal fields when using IF
Added the test cases in the above two bugs for regression
testing.
Added additional tests that demonstrate a incomplete fix.
Added a new factory method for Field_new_decimal to
create a field from an (decimal returning) Item.
In the new method made sure that all the precision and
length variables are capped in a proper way.
This is required because Item's can have larger precision
than the decimal fields and thus need to be capped when
creating a field based on an Item type.
Fixed the wrong typecast to Item_decimal.
When merging ranges during calculation of the result of OR
to two range sets the current range may be obsoleted by the
resulting merged range.
The first overlapping range can be obsoleted as well.
Fixed by moving the pointer to the first overlapping range to the
pointer of the resulting union range.
Added few comments at key places in key_or().
Problem: Some system functions that could return different values on
master and slave were not marked unsafe. In particular:
GET_LOCK
IS_FREE_LOCK
IS_USED_LOCK
MASTER_POS_WAIT
RELEASE_LOCK
SLEEP
SYSDATE
VERSION
Fix: Mark these functions unsafe.
Fixed a problem with the test case when executed with ps-protocol.
There the conflicing lock would be noticed during prepare, not
during execution of the insert - leading to a different (but
equally appropriate) error message.
DELETE IGNORE
The ER_CANT_UPDATE_USED_TABLE_IN_SF_OR_TRG error was set in the
diagnostics area when it happened, but the DELETE cleanup code
never checked for a non-fatal error condition, thus trying to
set diag.area to "ok". This triggered an assert checking that
the diag.area was empty.
The fix was to test if there existed a non-fatal error condition
(thd->is_error() before ok'ing the operation.
The problem was a "self-deadlock" if the connection issuing INSERT DELAYED
had both the global read lock (FLUSH TABLES WITH READ LOCK) and LOCK TABLES
mode active. The table being inserted into had to be different from the
table(s) locked by LOCK TABLES.
For INSERT DELAYED, the connection thread waits until the handler thread has
opened and locked its table before returning. But since the global read lock
was active, the handler thread would be unable to lock and would wait for the
global read lock to go away.
So the handler thread would be waiting for the connection thread to release
the global read lock while the connection thread was waiting for the handler
thread to lock the table. This gave a "self-deadlock" (same connection,
different threads).
The deadlock would only happen if we also had LOCK TABLES mode since the
INSERT otherwise will try to get protection against global read lock before
starting the handler thread. It will then notice that the global read lock
is owned by the same connection and report ER_CANT_UPDATE_WITH_READLOCK.
This patch removes the deadlock by reporting ER_CANT_UPDATE_WITH_READLOCK
also if we are inside LOCK TABLES mode.
Test case added to delayed.test.
WHERE conditions
check_group_min_max() checks if the loose index scan
optimization is applicable for a given WHERE condition, that is
if the MIN/MAX attribute participates only in range predicates
comparing the corresponding field with constants.
The problem was that it considered the whole predicate suitable
for the loose index scan optimization as soon as it encountered
a constant as a predicate argument. This is obviously wrong for
cases when a constant is the first argument of a predicate
which does not satisfy the above condition.
Fixed check_group_min_max() so that all arguments of the input
predicate are considered to decide if it passes the test, even
though a constant has already been encountered.
MySQL manual describes values of the YEAR(2) field type as follows:
values 00 - 69 mean 2000 - 2069 years and values 70 - 99 mean 1970 - 1999
years. MIN/MAX and comparison functions was comparing them as int values
thus producing wrong result.
Now the Arg_comparator class is extended with compare_year function which
performs correct comparison of the YEAR type.
The Item_sum_hybrid class now uses Item_cache and Arg_comparator objects to
correctly calculate its value.
To allow Arg_comparator to use func_name() function for Item_func and Item_sum
objects the func_name declaration is moved to the Item_result_field class.
A helper function is_owner_equal_func is added to the Arg_comparator class.
It checks whether the Arg_comparator object owner is the <=> function or not.
A helper function setup is added to the Item_sum_hybrid class. It sets up
cache item and comparator.
Post-push fix: Removed MTRv1 arguments according to the
original patch. Although there is a version check, the patch
was pushed to a 5.1 GA staging tree, while the version check
considers version 5.2. This makes the deprecated parameters
to be used, despite the fact that they are not valid anymore.
Part of MTRv1 is currently used in RQG semisync test, and this
was causing the test to fail on slave startup.
It should be safe to uncomment when merging up to celosia.
init_read_record() - (records.cc:274)
Item_cond::used_tables_cache was accessed in
records.cc#init_read_record() without being initialized. It had
not been initialized because it was wrongly assumed that the
Item's variables would not be accessed, and hence
quick_fix_field() was used instead of fix_fields() to save a few
CPU cycles at creation time.
The fix is to properly initilize the Item by replacing
quick_fix_field() with fix_fields().
One statement that have more than one different tables to update with
autoinc columns just was marked as unsafe in mixed mode, so the unsafe
warning can't be produced in statement mode.
To fix the problem, mark the statement as unsafe in statement mode too.
This patch borrows ideas, text and code from Kristofer
Pettersson's patch.
An assignment of a system variable sharing the same base
name as a declared stored procedure variable in the same
context could lead to a crash.
The reason was that during the parsing of the syntactic
rule 'option_value' an uninitialized set_var object was
pushed to the parameter stack of the SET statement. The
parent rule 'option_type_value' interpreted the existence
of variables on the parameter stack as an assignment and
wrapped it in a sp_instr_set object.
As the procedure later was executed an attempt was made
to run the method 'check()' on an uninitialized member
object (NULL value) belonging to the previously created
but uninitialized object.
This patch refactors the 'internal_variable_name' rule and
copies the semantic analysis part to the depending parent
rule: 'option_value'. This makes it possible to account
for any prefixes affecting the interpretation of the
internal_variable_name.
Until-pos guarding did not distiguish the master originated events from ones that the slave
can introduce to the relay log e.g Rotate to the next relay log at slave restarting.
The local Rotate's coordinate are incomparable with the Until-master-pos.
That led to the unexpectable stop this bug describes.
Fixed with to avoid Until-master-pos comparison for a local slave's event.
Notice that if --replicate-same-server is true such event is treated as coming from
the master side.
deadlock was encountered
The bug is caused by an inconsistent handling of the IGNORE
clause. A read from a const table caused a lock timeout
(ER_LOCK_TIMEOUT) in innodb. Since the IGNORE clause was
given, the timeout was converted into a warning instead of
an error, thus not populating the diagnostics area. When
innodb subsequently marked the transaction for rollback,
mysql asserted since the diag.area was empty.
This patch consists of only a test case, as the bug itself
was fixed by the patch for Bug #46539
on any access
Archive engine for 5.1 (and latter) version uses a modified
version of zlib (azlib). These two version are incompatible
so a proper upgrade is needed before tables created in 5.0
can be used reliable.
This upgrade can be performed using repair. But due to lack
of test its risky to allow upgrade for now. This patch addresses
only the crashing issue. Any attempt to repair will be blocked.
Eventually repair can be allowed to run through (which will also
cause an upgrade from older version to newer) but only after a
thorough testing.
The additional patch. That 'loadxml.test' failure was actually about our testing system,
not the code.
Firstly we need a new mysqltest command, wich i called 'send_eval'. So the expression
can be evaluated, then started in a parallel thread. We only have separane 'send' and
'eval' commands at the moment.
Then we need to add the waiting code after the 'KILL' to our test, so the thread will be killed
before the test goes further. The present 'reap' command doesn't handle the killed threads
well.
per-file comments:
client/mysqltest.cc
Bug#42520 killing load .. infile Assertion failed: ! is_set(), file .\sql_error.cc, line 8
The 'send_eval' command implemented.
mysql-test/r/loadxml.result
Bug#42520 killing load .. infile Assertion failed: ! is_set(), file .\sql_error.cc, line 8
test result updated.
mysql-test/t/loadxml.test
Bug#42520 killing load .. infile Assertion failed: ! is_set(), file .\sql_error.cc, line 8
test case added.
------------------------------------------------------------
revno: 2476.784.4
revision-id: sp1r-davi@moksha.local-20071008114751-46069
parent: sp1r-davi@moksha.local-20071003002731-48537
committer: davi@moksha.local
timestamp: Mon 2007-10-08 08:47:51 -0300
message:
Bug#27249 table_wild with alias: select t1.* as something
Aliases to table wildcards are silently ignored, but they should
not be allowed as it is non-standard and currently useless. There
is not point in having a alias to a wildcard of column names.
The solution is to rewrite the select_item rule so that aliases
for table wildcards are not accepted.
Contribution by Martin Friebe
------------------------------------------------------------
revno: 2597.4.17
revision-id: sp1r-davi@mysql.com/endora.local-20080328174753-24337
parent: sp1r-anozdrin/alik@quad.opbmk-20080328140038-16479
committer: davi@mysql.com/endora.local
timestamp: Fri 2008-03-28 14:47:53 -0300
message:
Bug#15192 "fatal errors" are caught by handlers in stored procedures
The problem is that fatal errors (e.g.: out of memory) were being
caught by stored procedure exception handlers which could cause
the execution to not be stopped due to a continue handler.
The solution is to not call any exception handler if the error is
fatal and send the fatal error to the client.
------------------------------------------------------------
revno: 3317
revision-id: davi.arnaut@sun.com-20090522170916-fzc5ca3tjs9roy1t
parent: patrick.crews@sun.com-20090522152933-ole8s3suy4zqyvku
committer: Davi Arnaut <Davi.Arnaut@Sun.COM>
branch nick: 41860-6.0
timestamp: Fri 2009-05-22 14:09:16 -0300
message:
Bug#41860: Without Windows named pipe
The problem was that the patch for Bug#10374 broke named pipe
and shared memory transports on Windows due to a failure to
implement a dummy poll method for transports other than BSD
sockets. Another problem was that mysqltest lacked support
for named pipe and shared memory connections, which lead to
misleading test cases that were supposed run common queries
over both transports.
The solution is to properly implement, at the VIO layer, the
poll and is_connected methods. The is_connected method is
implemented for every suppported transport and the poll one
only where it makes sense. Furthermore, support for named pipe
and shared memory connections is added to mysqltest as to
enable testing of both transports using the test suite.
------------------------------------------------------------
revno: 2597.37.3
revision-id: sp1r-davi@mysql.com/endora.local-20080328123626-16430
parent: sp1r-anozdrin/alik@quad.opbmk-20080327125300-11290
committer: davi@mysql.com/endora.local
timestamp: Fri 2008-03-28 09:36:26 -0300
message:
Bug#10374 GET_LOCK does not let connection to close on the server side if it's aborted
The problem is that the server doesn't detect aborted connections which
are waiting on a lock or sleeping (user sleep), wasting system resources
for a connection that is already dead.
The solution is to peek at the connection every five seconds to verify if
the connection is not aborted. A aborted connection is detect by polling
the connection socket for available data to be read or end of file and in
case of eof, the wait is aborted and the connection killed.
