Bug #18651 Problem in test "kill": Server crash (different)
Submitted: 30 Mar 2006 13:37 Modified: 21 Jun 2006 20:22
Reporter: Joerg Bruehe Email Updates:
Status: Can't repeat Impact on me:
None 
Category:MySQL Server Severity:S1 (Critical)
Version:5.0.20 OS:MacOS (OS X (x86))
Assigned to: Chad MILLER CPU Architecture:Any

[30 Mar 2006 13:37] Joerg Bruehe
Description:
(Release ?) build of 5.0.20, code of 2006-Mar-29/30.

This error differs from #18650; 
it occurs in the same build, also on OS X x86, but in the PS-protocol run.

It looks very similar to the last entries in #9712 which is said to be fixed ...

===== OS X, x86, "max", PS protocol =====
kill                           [ fail ]

Errors are (from /Users/mysqldev/osx-imac-x86/test/mysql-max-5.0.20-osx10.4-i686/mysql-test/var/log/mysqltest-time) :
mysqltest: At line 42: query 'select ((@id := kill_id) - kill_id) from t1' failed: 2013: Lost connection to MySQL server during query
(the last lines may be the most important ones)
Result from queries before failure can be found in r/kill.log

Killing Possible Leftover Processes
Restoring snapshot of databases
Resuming Tests
===== end

How to repeat:
Run the test suite, PS protocol.
[19 Apr 2006 8:02] Magnus BlÄudd
This is a classical problem with _when_ we detect if the connection to the server is still connected or not. If we do it before we send anything to the server, we can safely say that server is gone and set error 2006. If we have to send something and don't get a reply, we will set error 2013. Since the implementation of TCP sockets differs and we can use sockets, named pipes shared mem etc to connect to the server we must use a little different techniques to detect the error early.

There are other bugs related. See for example bug#2845 which is the "mother"  bug.
[21 Jun 2006 18:48] Chad MILLER
I can't reproduce with debug flags on.  Trying without.