~mepsql-committers/mepsql/mepsql-5.1 : changes

~mepsql-committers/mepsql/mepsql-5.1 » Changes from revision 3630

From Revision 3629 to 3610

Rev	Summary	Authors	Date
3629	Modified the AC_INIT version string and server name to MepSQ... Modified the AC_INIT version string and server name to MepSQL and alpha2.	Henrik Ingo	13 years ago
3628	Merge Makefile additions from mepsql-5.1-hack-make. Now this... Merge Makefile additions from mepsql-5.1-hack-make. Now this builds successfully.	Henrik Ingo	13 years ago
3627	Try some hack again, put back --without-plugin-innobase Try some hack again, put back --without-plugin-innobase	Henrik Ingo	13 years ago
3626	Try removing the second --without-plugin-innobase (compile-d... Try removing the second --without-plugin-innobase (compile-dist already does it once).	Henrik Ingo	13 years ago
3625	--with-partition doesn't work in newer autoconf versions any... --with-partition doesn't work in newer autoconf versions anymore. (Note, it is enabled with --with-plugins=partition,...)	Henrik Ingo	13 years ago
3624	Ok, also need to use Mark's configure options for make dist. Ok, also need to use Mark's configure options for make dist. https://answers.launchpad.net/mysqlatfacebook/+question/144091/+confirm?answer_id=0	Henrik Ingo	13 years ago
3623	Merge from launchpad before push. (These are the new changes... Merge from launchpad before push. (These are the new changes from mysqlatfacebook.)	Henrik Ingo	13 years ago
3622	Use same configure options as Mark uses for mysqlatfacebook ... Use same configure options as Mark uses for mysqlatfacebook internally. https://answers.launchpad.net/mysqlatfacebook/+question/144091/+confirm?answer_id=0	Henrik Ingo	13 years ago
3621	merge compile/bintar addition from mepsql/mysql-5.1 merge compile/bintar addition from mepsql/mysql-5.1	Henrik Ingo	13 years ago
3620	Allow --mysqlfork="" to be set also via "export MYSQLFORK". ... Allow --mysqlfork="" to be set also via "export MYSQLFORK". This allows for better backwards compatibility, as old versions of this script will just silently ignore it.	Henrik Ingo	13 years ago
3619	Add support for --mysqlfork=mepsql to scripts/make_binary_di... Add support for --mysqlfork=mepsql to scripts/make_binary_distribution.sh Default is good old "mysql", but this allows bakery scripts to produce differently branded TAR.	Henrik Ingo	13 years ago
3618	With this change the following will happen: With this change the following will happen: If a .ibd file points to a non-existing path, then the table corresponding to that .ibd file is created in that location. After alter table add column, add index, delete column, rename table operations the symlinks are still preserved.	Mark Callaghan	13 years ago
3617	Add information_schema.index_statistics. It has columns simi... Add information_schema.index_statistics. It has columns similar to table_statistics, but the data is counted per index. desc information_schema.index_statistics; Field Type Null Key Default Extra TABLE_SCHEMA varchar(192) NO TABLE_NAME varchar(192) NO INDEX_NAME varchar(192) NO TABLE_ENGINE varchar(192) NO ROWS_INSERTED bigint(21) NO 0 ROWS_UPDATED bigint(21) NO 0 ROWS_DELETED bigint(21) NO 0 ROWS_READ bigint(21) NO 0 ROWS_REQUESTED bigint(21) NO 0 ROWS_INDEX_FIRST bigint(21) NO 0 ROWS_INDEX_NEXT bigint(21) NO 0 IO_READ_BYTES bigint(21) NO 0 IO_READ_REQUESTS bigint(21) NO 0 IO_READ_SVC_USECS bigint(21) NO 0 IO_READ_SVC_USECS_MAX bigint(21) NO 0 IO_READ_WAIT_USECS bigint(21) NO 0 IO_READ_WAIT_USECS_MAX bigint(21) NO 0 IO_READ_OLD_IOS bigint(21) NO 0 The index names reported can lag reality after add/drop index. The "flush statistics" command will reset all counters and force the index names to be fixed. Otherwise, they are fixed when the number of indexes for a table changes. So, they are fixed automatically after drop index. There is no problem after add index. But the sequence drop index, ..., add index can be a problem because the number of indexes has not changed. The "problem" is limited to 1) counters are not reset and 2) the reported name might be wrong. I will eventually fix this, but right now the implementation is a very minor and simple change beyond what table_stats requires. This also fixes a serious but unlikely bug in get_table_stats -- mutex was not unlocked on some errors.	Mark Callaghan	13 years ago
3616	Add debug assert to find incorrect uses of c_ptr Add debug assert to find incorrect uses of c_ptr See http://bugs.mysql.com/bug.php?id=48053	Mark Callaghan	13 years ago
3615	Implement innodb_fast_free_list option to make free-list man... Implement innodb_fast_free_list option to make free-list management faster This adds the dynamic my.cnf parameter innodb_fast_free_list that provides an alternative to buf_flush_free_margin to flush dirty pages near the end of the LRU. This is done as they will soon be moved to the free list but only clean pages can be moved. The combination of buf_flush_free_margin and buf_flush_batch makes one pass of the tail of the LRU to count dirty pages and then one pass from the tail per page flushed. The buffer pool mutex is held during each pass and that mutex already has a lot of contention. For workloads that frequently move pages from the LRU to the free list (buffer pool pressure) this function is a bottleneck. The alternative uses one pass of the tail to both count the dirty pages and save the space & page ID values for them. The buffer pool mutex is held for that pass. Then it iterates on the array and flushes one page at a time via buf_flush_try_neighbors with the 'flush-neighbors' flag set to the value of the dynamic my.cnf parameter innodb_flush_neighbors_for_lru. The analysis above about the number of passes is the worst-case when innodb_flush_neighbors_for_lru is OFF or when there are few dirty pages per extent. Otherwise, mutiple dirty pages can be flushed per pass. More performance tests will be run but this is disabled by default and should be safe to add.	Mark Callaghan	13 years ago
3614	buf0buddy.c: new gcc optimizer changes where valgrind catche... buf0buddy.c: new gcc optimizer changes where valgrind catches intentionally undefined memory to earlier the before, move the valgrind ignore commands earlier. page0zip.c: zlib has known valgrind warnings, mark memory returned from our zlib memory allocator as defined to avoid it.	Mark Callaghan	13 years ago
3613	fix compilation warnings fix compilation warnings	Mark Callaghan	13 years ago
3612	Fix diffs for information_schema.index_stats Fix diffs for information_schema.index_stats	Mark Callaghan	13 years ago
3611	fix bad merges fix bad merges	Mark Callaghan	13 years ago
3610	Revert "Get fix for bug 54790 to reduce overhead from fcntl ... Revert "Get fix for bug 54790 to reduce overhead from fcntl on network recdeive" This broke ssl support and the real fix should be in 5.5	Mark Callaghan	13 years ago

Older »