Copy any data (currently table name and table index) that may be
destroyed after releasing the kernel mutex into internal cache's
storage.
This is done in efficient manner using ha_storage type and a given
string is copied only once into the cache's storage. Later additions of
the same string use the already stored string, thus allocating memory
only once per unique string.
Approved by: Marko
Add a type that stores chunks of data in its own storage and avoids
duplicates. Supported methods:
ha_storage_create()
Allocates new storage object.
ha_storage_put()
Copies a given data chunk into the storage and returns pointer to the
copy. If the data chunk is already present, a pointer to the existing
object is returned and the given data chunk is not copied.
ha_storage_empty()
Clears (empties) the storage from all data chunks that are stored in it.
ha_storage_free()
Destroys a storage object. Opposite to ha_storage_create().
Approved by: Marko
row_merge(): Add the assertion ut_ad(half > 0).
row_merge_sort(): Compute the half of the merge file correctly. The
previous implementation used truncating division, which may result in
loss of records when the file size in blocks is not a power of 2.
Use the newly introduced mem_alloc2() to use the memory that has been
allocated in addition to the requested memory. This is done in order to
avoid wasting memory.
Do not calculate the sizes and offsets of the chunks in advance in
table_cache_init() because it is unknown how much bytes will actually
be allocated by mem_alloc2(). Rather calculate these on the run: after
each chunk is allocated set its size and the offset of the next chunk.
Similar patch approved by: Marko
Some bug still remains, because innodb-index.test will lose some
records from the clustered index after add primary key (a,b(255),c(255))
when row_merge_block_t is reduced to 8192 bytes.
row_merge(): Add the parameter "half". Add some Valgrind instrumentation.
Note that either stream can end before the other one.
row_merge_sort(): Calculate "half" for row_merge().
mem_alloc2(): New macro. This is a variant of mem_alloc() that
returns the allocated size, which is equal to or greater than
the requested size.
mem_alloc_func(): Add the output parameter *size for the allocated size.
When it is set, adjust the parameter passed to mem_heap_alloc().
rec_copy_prefix_to_buf_old(), rec_copy_prefix_to_buf(): Use mem_alloc2()
instead of mem_alloc().
was actually obtained from the buddy allocator. This should avoid some
internal memory fragmentation in mem_heap_create() and mem_heap_alloc().
mem_area_alloc(): Change the in parameter size to an in/out parameter.
Adjust the size based on what was obtained from pool->free_list[].
mem_heap_create_block(): Adjust block->len to what was obtained from
mem_area_alloc().
rec_print_comp(): New function, sliced from rec_print_new().
rec_print_old(), rec_print_comp(): Print the untruncated length of the column.
row_merge_print_read, row_merge_print_write, row_merge_print_cmp:
New flags, to enable debug printout in UNIV_DEBUG builds.
row_merge_tuple_print(): New function for UNIV_DEBUG builds.
row_merge_read_rec(): Obey row_merge_print_read.
row_merge_buf_write(), row_merge_write_rec_low(),
row_merge_write_eof(): Obey row_merge_print_write.
row_merge_cmp(): Obey row_merge_print_cmp.
in fast index creation.
row_merge_write_eof(), row_merge_buf_write(): When UNIV_DEBUG_VALGRIND
is defined, fill the rest of the block (after the end-of-block marker)
with 0xff.
innodb_lock_waits. See
https://svn.innodb.com/innobase/InformationSchema/TransactionsAndLocks
for design notes.
Things that need to be resolved before this goes live:
* MySQL must add thd_get_thread_id() function to their code
http://bugs.mysql.com/30930
* Allocate memory from mem_heap instead of using mem_alloc()
* Copy table name and index name into the cache because they may be
freed later which will result in referencing freed memory
Approved by: Marko
row_merge_read_rec(): Correct a typo in a comment. Fix error in
arithmetics when the record spans two blocks.
row_merge_write_rec_low(): Add a "size" parameter. Add debug assertions
about extra_size and size.
row_merge_write_rec(): After writing a record, properly advance the
buffer pointer.
that row_merge_blocks() will have some work to do when
row_merge_block_t is shrunk to 8192 bytes.
Currently, this will cause a debug assertion failure, because
row_merge_cmp() is considering all columns, not just the unique ones.
existed before the fast-index-creation code was merged, in r1414.
Do not call row_prebuilt_table_obsolete(innodb_table) until after
the table has been renamed to a temporary name. Otherwise, the
following could happen when creating a clustered index on table T:
(a1) copy T to T#1
(a2) mark T obsolete
(b1) add the name of T to the background drop list
(a3) rename T to T#2 and T#1 to T
(b2) drop the table on the background
These steps are executed by two threads, a and b.
If b1 is executed after a3, all is fine: the name T#2 will be added
to the background drop list. If b1 is executed before a3, the name
T will be added, and T#2 will survive.
row_add_table_to_background_drop_list(): Replace the dict_table_t*
parameter with const char* (the table name).
row_check_table_for_mysql(): Replace prebuilt->table with table.
fast index creation.
innobase_convert_tablename(): New function, to remove the filename
encoding of a database and table name.
ha_innobase::add_index(): Invoke innobase_convert_tablename() before
calling my_error().
innodb-index.result: Update the error messages.
UNIV_MEM_ASSERT_RW(): New macro, to check that the contents of a memory
area is defined.
UNIV_MEM_ASSERT_W(): New macro, to check that a memory area is writable.
UNIV_MEM_ASSERT_AND_FREE(): New macro, to check that the memory is
writable before declaring it free (unwritable). This replaces UNIV_MEM_FREE()
in many places.
mem_init_buf(): Declare the memory undefined.
mem_erase_buf(): Declare the memory freed.
trx_t: Remove dict_undo_list and dict_redo_list.
innobase_create_temporary_tablename(): Replace TEMP_TABLE_PREFIX with
a table name suffix "#1" or "#2". In this way, the user can restore
precious data, should anything go wrong. It is possible to reach an
inconsistent state, because the creation, deletion and renaming of
single-table tablespaces are not transactional.
ut_print_namel(), fil_make_ibd_name(), innobase_rename_table(): Remove
the special treatment of TEMP_TABLE_PREFIX.
Introduce TEMP_INDEX_PREFIX == 0xff for temporary indexes. This byte
cannot occur in index names since MySQL 4.1. However, it might have
been possible to use this byte in MySQL 4.0.
recv_recovery_from_checkpoint_finish(): Call the new function
row_merge_drop_temp_indexes(), to drop all indexes whose name starts
with the byte 0xff.
row_merge_rename_indexes(): Renamed from row_merge_rename_index().
Remove the parameter "index".
row_drop_table_for_mysql(): Unconditionally call trx_commit_for_mysql().
row_drop_table_for_mysql_no_commit(): Correct the function commit,
based on the corrected comment of row_drop_table_for_mysql(). Rely on
table->to_be_dropped instead of TEMP_TABLE_PREFIX.
ha_innobase::add_index(): Simplify the control flow.
is inline and accesses lock_t members.
As advised by Marko rename lock_get_type() to lock_get_type_low() and
create a public non-inline function lock_get_type() which calls
lock_get_type_low().
Approved by: Marko
