>>>>> "Joe" == Joe Kislo <kislo@stripped> writes:
Joe> Michael Widenius wrote:
>> The above problem is probably something that is very data dependent;
>> To fix and find this quickly I would need a copy of the tables in
>> If possible, please ftp them, and a copy of this mail, to
>> ftp://www.mysql.com/pub/mysql/secret and I will take a look at this
>> before the next MySQL release!
Joe> Monty, I never got a response to the bug I reported. Somebody on the
Joe> list 'confirmed it' but I don't think I ever got a response from any
Joe> mysql engineers. I realize that the company I am doing this work for
Joe> does not have a mysql subscription; so I understand if it's not a
Joe> priority [although it DOES crash the stable version of mysql]. However
Joe> I need to tell managment -something-, even if it's been confirmed and
Joe> their fixing it, otherwise the word will come down to dump mysql and go
Joe> with informix or oracle. Will this be fixed in the next build, and when
Joe> will that be? Nobody here is particularly happy about the database
Joe> server crashing.
Joe> Here's part of the original bug report: you can get the whole thing if
Joe> you scroll back on the list under the subject: "Mysqlserver crashes
Joe> [EG:coredump] with drop table and locks.
Joe> If you have two disjoint processes aquire a write table lock on the same
Joe> table, and one of them drops the table from the database, the mysql
Joe> server crashes
Joe> I have been able to repeat this on 3.22.25 and 3.22.27, on both x86
Joe> 2.2.x and sparc solaris 7.
Joe> Create two tables
Joe> establish two mysql connections [in different windows]
Joe> aquire write lock on both tables in both windows [using 1 sql
Joe> one will get the locks, and one will have to wait
Joe> have the one which got the lock, drop one of the tables it has a
Joe> lock on
Joe> the Mysql server will immediatly crash, and both connections
Joe> will be
Joe> None, server crashes. It's corrupted my tables a couple times
Sorry for not going back to you, but I haven't had time to follow the
MySQL mailing list for 4 weeks because I have been busy getting the
last changes to MySQL 3.23 to get it 'stable' as soon as possible.
At least MySQL 3.23.7 (the version I am working on) doesn't have this
problem. As this problem is quite easy to avoid (as long as one knows
about it) is shouldn't be really critical to fix this as 3.23.6 should
be usable very soon.