This bug was discovered previously in another scenario as MySQL Bug#42916.

When executing OPTIMIZE TABLE after a kill -9 recovery, maria asserted as follows:

mysqld: ma_pagecache.c:3279: pagecache_read: Assertion `pageno < ((1ULL) << 40)' failed.

backtrace:

#8 0x00133de8 in __assert_fail () from /lib/libc.so.6
#9 0x08533d59 in pagecache_read (pagecache=0x9264d00, file=0xa5b2424, pageno=1814651154228852, level=3, buff=0x99a08ef0 "", type=PAGECACHE_LSN_PAGE,
lock=PAGECACHE_LOCK_LEFT_UNLOCKED, page_link=0x99a08e80) at ma_pagecache.c:3279
#10 0x085529fa in _ma_fetch_keypage (page=0x99a0af34, info=0xa5b6e78, keyinfo=0xa5b285c, pos=14865622255442755584, lock=PAGECACHE_LOCK_LEFT_UNLOCKED,
level=3, buff=0x99a08ef0 "", return_buffer=0 '\000') at ma_page.c:108
#11 0x08597291 in sort_one_index (param=0x99a0dde4, info=0xa5b6e78, keyinfo=0xa5b285c, pagepos=8589934595, new_file=41) at ma_check.c:3112
#12 0x0859738a in sort_one_index (param=0x99a0dde4, info=0xa5b6e78, keyinfo=0xa5b285c, pagepos=616182317536301164, new_file=41) at ma_check.c:3132
#13 0x08596d0d in maria_sort_index (param=0x99a0dde4, info=0xa5b6e78, name=0x99a0db17 "./test/table10000_maria_key_pk_parts_2_varchar_1024_not_null#P#p0")
at ma_check.c:3006
#14 0x0853c109 in ha_maria::repair (this=0xa5965f8, thd=0xa498878, param=0x99a0dde4, do_optimize=true) at ha_maria.cc:1486
#15 0x0853b710 in ha_maria::optimize (this=0xa5965f8, thd=0xa498878, check_opt=0xa49a4e8) at ha_maria.cc:1365
#16 0x083bee33 in handler::ha_optimize (this=0xa5965f8, thd=0xa498878, check_opt=0xa49a4e8) at handler.cc:3244
#17 0x083c4d78 in handle_opt_part (thd=0xa498878, check_opt=0xa49a4e8, file=0xa5965f8, flag=1) at ha_partition.cc:963
#18 0x083c52ee in ha_partition::handle_opt_partitions (this=0xa596160, thd=0xa498878, check_opt=0xa49a4e8, flag=1) at ha_partition.cc:1114
#19 0x083c4b4a in ha_partition::optimize (this=0xa596160, thd=0xa498878, check_opt=0xa49a4e8) at ha_partition.cc:873
#20 0x083bee33 in handler::ha_optimize (this=0xa596160, thd=0xa498878, check_opt=0xa49a4e8) at handler.cc:3244
#21 0x083e3ad8 in mysql_admin_table (thd=0xa498878, tables=0xa57b478, check_opt=0xa49a4e8, operator_name=0x88943af "optimize", lock_type=TL_WRITE,
open_for_modify=false, no_warnings_for_error=false, extra_open_options=0, prepare_func=0, operator_func=
(int (handler::)(handler *, THD *, HA_CHECK_OPT *)) 0x83bee00 <handler::ha_optimize(THD, HA_CHECK_OPT*)>, view_operator_func=0) at sql_table.cc:4828
#22 0x083e4df4 in mysql_optimize_table (thd=0xa498878, tables=0xa57b478, check_opt=0xa49a4e8) at sql_table.cc:5102
#23 0x0828d634 in mysql_execute_command (thd=0xa498878) at sql_parse.cc:3118
#24 0x08296857 in mysql_parse (thd=0xa498878, inBuf=0xa57b250 "OPTIMIZE TABLE `test`.`table10000_maria_key_pk_parts_2_varchar_1024_not_null`", length=77,
found_semicolon=0x99a28228) at sql_parse.cc:6079
#25 0x08288b11 in dispatch_command (command=COM_QUERY, thd=0xa498878, packet=0xa55b531 "", packet_length=77) at sql_parse.cc:1253
#26 0x08287d4f in do_command (thd=0xa498878) at sql_parse.cc:891
#27 0x08284e96 in handle_one_connection (arg=0xa498878) at sql_connect.cc:1599
#28 0x00a08919 in start_thread () from /lib/libpthread.so.0
#29 0x001ede5e in clone () from /lib/libc.so.6

Before the OPTIMIZE TABLE, the table was checked with CHECK TABLE EXTENDED and ANALYZE TABLE. None reported a corruption when they probably should have.

bzr version-info:

revision-id: <email address hidden>
date: 2010-06-14 15:17:54 +0400
build-date: 2010-07-08 00:50:09 -0700
revno: 2789
branch-nick: maria

The tablespace before recovery + core + binary will be uploaded shortly.