MySQL Bugs: #35733: main.symlink.test fails

Bug #35733	main.symlink.test fails
Submitted:	1 Apr 2008 9:01	Modified:	20 May 2008 23:14
Reporter:	Alexander Nozdrin	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server: Tests	Severity:	S7 (Test Cases)
Version:	6.0	OS:	Any
Assigned to:	Dmitry Lenev	CPU Architecture:	Any

Description:
main.symlink.test fails.

How to repeat:
https://intranet.mysql.com/secure/pushbuild/xref.pl?testname=main.symlink

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/44972

ChangeSet@1.2620, 2008-04-07 11:53:00+04:00, dlenev@mockturtle.local +1 -0
  Tentative fix for bug #35733 "main.symlink.test fails".
  
  The problem was that symlink.test failed due to debug assertions
  or due to errors emitted for wrong statements (in non-debug builds)
  if server was run with --thread-handling=pool-of-threads option.
  
  The above problems were caused by that on the one hand code mi_create()
  which handles case when error is returned relies on my_errno being set
  to appropriate non-0 value, but on the other it has not done this
  in case when error was emitted due to discovery that share for table
  which we are going to create is already present in the list of open
  shares. Running server with pool-of-threads option just exposed this
  bug since in this scheduler mode my_errno is reset to 0 when connection
  (THD) is reattached to the "real" thread.
  
  This fix ensures that code in mi_create() properly sets my_errno in
  the case described above.
  
  Question to reviewers are marked by QQ.
  
  QQ: Does it make sense to push this fix to 5.0/5.1 (IMO yes)?

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/44977

ChangeSet@1.2584, 2008-04-07 12:59:57+04:00, dlenev@mockturtle.local +1 -0
  Fix for bug #35733 "main.symlink.test fails".
  
  The problem was that symlink.test failed due to debug assertions or
  due to errors emitted for wrong statements (in non-debug builds)
  if 6.0 version of server was run with --thread-handling=pool-of-threads
  option.
  
  The above problems were caused by that on the one hand code mi_create()
  which handles case when error is returned relies on my_errno being set
  to appropriate non-0 value, but on the other it has not done this
  in case when error was emitted due to discovery that share for table
  which we are going to create is already present in the list of open
  shares. Running server with pool-of-threads option just exposed this
  bug since in this scheduler mode my_errno is reset to 0 when connection
  (THD) is reattached to the "real" thread.
  
  This fix ensures that code in mi_create() properly sets my_errno in
  the case described above.
  
  Since original reason that caused this bug to be exposed in 6.0 tree
  might also under very special conditions lead to incorrect behavior
  of earlier versions of server (i.e. to errors emitted for wrong
  statements) this patch is also being pushed into 5.1 tree.
  
  No test case is present as this code is already covered by a test
  case + pushbuild options combination in 6.0.

Pushed into 6.0.6-alpha

Noted in 6.0.6 changelog.

An incorrect error and message was produced for attempts to create a
MyISAM table with an index (.MYI) filename that was already in use by
some other MyISAM table that was open at the same time. For example,
this might happen if you use the same value of the INDEX DIRECTORY
table option for tables belonging to different databases.

Resetting report to PQ waiting for push into 5.1.x.

Pushed into 5.1.25-rc

Noted in 5.1.25 changelog.

Pushed into 5.1.25-rc  (revid:sp1r-kostja@bodhi.(none)-20080520073817-17550) (version source revid:sp1r-kostja@bodhi.(none)-20080520073817-17550) (pib:3)