Description:
With 5.1-performance-version, latest revision:
mikael@mysql.com-20090702155353-v2xbu1tlct7epb95
./mtr rpl_trigger --valgrind
I get, on top of BUG#46081 and BUG#45630, this one:
==31105== by 0x97406A: mem_heap_create_block (mem0mem.c:362)
==31105== by 0x9733A4: mem_heap_create_func (mem0mem.ic:421)
==31105== by 0x973523: mem_alloc_func (mem0mem.ic:511)
==31105== by 0x9A8B4E: thr_local_create (thr0loc.c:166)
==31105== by 0x9A8A29: thr_local_get (thr0loc.c:78)
==31105== by 0x9A8B08: thr_local_get_in_ibuf_field (thr0loc.c:146)
==31105== by 0x95E593: ibuf_enter (ibuf0ibuf.c:251)
==31105== by 0x95E95F: ibuf_data_init_for_space (ibuf0ibuf.c:487)
==31105== by 0x9511C7: fil_ibuf_init_at_db_start (fil0fil.c:1488)
==31105== by 0x95E79B: ibuf_init_at_db_start (ibuf0ibuf.c:402)
==31105== by 0x9C21FE: dict_boot (dict0boot.c:384)
==31105== by 0x9C22C0: dict_create (dict0boot.c:422)
==31105== by 0x9A5A59: innobase_start_or_create_for_mysql (srv0start.c:1502)
==31105== by 0x9293F1: _ZL13innobase_initPv (ha_innodb.cc:1785)
==31105==
==31105==
==31105== 1,753,184 bytes in 120 blocks are still reachable in loss record 8 of 8
==31105== at 0x4C265AE: malloc (vg_replace_malloc.c:207)
==31105== by 0x974DD5: mem_area_alloc (mem0pool.c:334)
==31105== by 0x97406A: mem_heap_create_block (mem0mem.c:362)
==31105== by 0x9733A4: mem_heap_create_func (mem0mem.ic:421)
==31105== by 0x973523: mem_alloc_func (mem0mem.ic:511)
==31105== by 0x9A0202: srv_init (srv0srv.c:963)
==31105== by 0x9A100B: srv_boot (srv0srv.c:1511)
==31105== by 0x9A52E4: innobase_start_or_create_for_mysql (srv0start.c:1209)
==31105== by 0x9293F1: _ZL13innobase_initPv (ha_innodb.cc:1785)
==31105== by 0x7FFEEE: ha_initialize_handlerton(st_plugin_int*) (handler.cc:433)
==31105== by 0x8BAC3D: _ZL17plugin_initializeP13st_plugin_int (sql_plugin.cc:1003)
==31105== by 0x8BE8FD: plugin_init(int*, char**, int) (sql_plugin.cc:1216)
==31105== by 0x6B9869: _ZL22init_server_componentsv (mysqld.cc:3973)
==31105== by 0x6BA308: main (mysqld.cc:4445)
I don't get this with 5.1-main, so maybe this is summit-specific.
This isn't visible in pushbuild2 for 5.1-performance-version, likely because the testsuite aborts before rpl_trigger in mixed mode due to errors in zlib. But it is visible in the tree which contains the merge of summit into azalea, and not visible in azalea, so this really looks summit-specific.
I'm using valgrind-3.3.1-Debian.
How to repeat:
run the test.
