MySQL Bugs: #50558: show full columns + flush table crashes

Bug #50558	show full columns + flush table crashes
Submitted:	22 Jan 2010 20:26	Modified:	30 Dec 2014 19:10
Reporter:	Shane Bester (Platinum Quality Contributor)	Email Updates:
Status:	Can't repeat	Impact on me:	None
Category:	MySQL Server: Locking	Severity:	S1 (Critical)
Version:	5.5.99-m3-debug	OS:	Windows (xp)
Assigned to:		CPU Architecture:	Any

Description:
I branched mysql-next-mr-bugfixing and built debug server that crashes

mysqld.exe!Field_varstring::sql_type()[field.cc:7038]
mysqld.exe!get_schema_column_record()[sql_show.cc:3985]
mysqld.exe!fill_schema_show_cols_or_idxs()[sql_show.cc:2950]
mysqld.exe!get_all_tables()[sql_show.cc:3267]
mysqld.exe!get_schema_tables_result()[sql_show.cc:6519]
mysqld.exe!JOIN::exec()[sql_select.cc:1796]
mysqld.exe!mysql_select()[sql_select.cc:2468]
mysqld.exe!handle_select()[sql_select.cc:271]
mysqld.exe!execute_sqlcom_select()[sql_parse.cc:4953]
mysqld.exe!mysql_execute_command()[sql_parse.cc:2135]
mysqld.exe!sp_instr_stmt::exec_core()[sp_head.cc:2952]
mysqld.exe!sp_lex_keeper::reset_lex_and_exec_core()[sp_head.cc:2774]
mysqld.exe!sp_instr_stmt::execute()[sp_head.cc:2889]
mysqld.exe!sp_head::execute()[sp_head.cc:1261]
mysqld.exe!sp_head::execute_procedure()[sp_head.cc:2005]
mysqld.exe!mysql_execute_command()[sql_parse.cc:4291]
mysqld.exe!mysql_parse()[sql_parse.cc:5989]
mysqld.exe!dispatch_command()[sql_parse.cc:1140]
mysqld.exe!do_command()[sql_parse.cc:810]
mysqld.exe!do_handle_one_connection()[sql_connect.cc:1172]
mysqld.exe!handle_one_connection()[sql_connect.cc:1112]
mysqld.exe!pfs_spawn_thread()[pfs.cc:1011]
mysqld.exe!pthread_start()[my_winthread.c:61]
mysqld.exe!_callthreadstartex()[threadex.c:348]
mysqld.exe!_threadstartex()[threadex.c:331]
kernel32.dll!FlsSetValue()

Exact revision was this:
[sbester@box1 mysql-next-mr-bugfixing]$ bzr log |head -n 20
------------------------------------------------------------
revno: 2972
committer: Alexander Nozdrin <alik@sun.com>
branch nick: mysql-next-mr-bugfixing
timestamp: Fri 2010-01-22 11:20:13 +0300
message:
  Auto-merge from mysql-next-mr.

How to repeat:
start mysqld with --performance-schema=1

concurrency bug.  in one thread flush a table. in another thread show full columns:

delimiter $
drop procedure if exists `p1` $
create procedure `p1`(`a` int)
begin
  declare `i` int default '0';
  repeat
    show full columns from `performance_schema`.`events_waits_current`;
    flush table `performance_schema`.`events_waits_current`;
    set `i` = `i` + 1;
  until `i` > `a` end repeat;

end $

delimiter ;

Now, start two or more connections and run this SP:

1) mysql -uroot test -e "call p1(10000)" > out1.txt
2) mysql -uroot test -e "call p1(10000)" > out.txt

sorry, this is NOT performance_schema related.  It seems 5.5 is just more broken than I thought.  I repeated it with normal tables also, in 3+ threads (was a bit harder):

delimiter $
drop table if exists t1 $
create table t1(a varchar(255))engine=memory$
insert into t1 values (),(),()$
drop procedure if exists `p1` $
create procedure `p1`(`a` int)
begin
  declare `i` int default '0';
  repeat
    show full columns from `t1`;
    flush table `t1`;
    set `i` = `i` + 1;
  until `i` > `a` end repeat;

end $

delimiter ;

I couldn't repeat on Windows Vista X64 (using older tables it crashed with drop procedure..command) however with fresh tables so far I was not able to repeat:

:\mysql>bin\mysqld --standalone --console --performance-schema=1
00125  9:47:18 [Note] Buffered information: Performance schema enabled.

00125  9:47:18 [Note] Plugin 'FEDERATED' is disabled.
nnoDB: The InnoDB memory heap is disabled
nnoDB: Mutexes and rw_locks use Windows interlocked functions
nnoDB: The first specified data file .\ibdata1 did not exist:
nnoDB: a new database to be created!
00125  9:47:18  InnoDB: Setting file .\ibdata1 size to 10 MB
nnoDB: Database physically writes the file full: wait...
00125  9:47:18  InnoDB: Log file .\ib_logfile0 did not exist: new to be created
nnoDB: Setting log file .\ib_logfile0 size to 5 MB
nnoDB: Database physically writes the file full: wait...
00125  9:47:19  InnoDB: Log file .\ib_logfile1 did not exist: new to be created
nnoDB: Setting log file .\ib_logfile1 size to 5 MB
nnoDB: Database physically writes the file full: wait...
nnoDB: Doublewrite buffer not found: creating new
nnoDB: Doublewrite buffer created
nnoDB: Creating foreign key constraint system tables
nnoDB: Foreign key constraint system tables created
00125  9:47:19 InnoDB Plugin 1.0.6 started; log sequence number 0
00125  9:47:20 [Note] Event Scheduler: Loaded 0 events
00125  9:47:20 [Note] bin\mysqld: ready for connections.
ersion: '5.5.99-m3-Win X64-debug'  socket: ''  port: 3306  Source distribution

try call p1(1000000)  and 20 threads then .... or try performance_schema tables, they seemed easier to crash server than 'normal' tables.

Not repeatable on Linux for me with mysql-next-mr-bugfixing revision 2976

Please, check if this is still repeatable on current mysql-trunk and/or mysql-next-mr trees in your environment.

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

Cannot repeat on 5.6.20, so closing this bug.