Re: MySQL crashes on amd64

From: Greg 'groggy' Lehey <grog_at_FreeBSD.org>
Date: Sun, 6 Nov 2005 08:27:57 +1030
On Saturday,  5 November 2005 at 20:56:10 +0100, Stefan Cars wrote:

> Kris Kennaway wrote:
>> On Wed, Nov 02, 2005 at 10:38:43AM +1030, Greg 'groggy' Lehey wrote:
>>> On Monday, 31 October 2005 at 11:41:21 +0100, Stefan Cars wrote:
>>>> Kris Kennaway wrote:
>>>>> On Sun, Oct 30, 2005 at 03:52:42PM -0500, Daniel Eischen wrote:
>>>>>> On Sun, 30 Oct 2005, Kris Kennaway wrote:
>>>>>>> On Sun, Oct 30, 2005 at 02:45:57PM +0100, Stefan Cars wrote:
>>>>>>>
>>>>>>>> No. I don't need both. Removed libpthread.so.1 and still the same
>>>>>>>> problem.
>>>>>>>
>>>>>>> You removed the mapping entry, or the file itself?  Daniel asked for
>>>>>>> the latter.
>>>>>>
>>>>>> Or perhaps try using libmap.conf to map libpthread.so.1 to
>>>>>> libpthread.so.2 without any other mappings:
>>>>>>
>>>>>>	libpthread.so.1		libpthread.so.2
>>>>>
>>>>> This is just a partial workaround, of course, and any of your
>>>>> binaries that are also linked to two versions of other libraries
>>>>> may still experience crashes (or may start to crash later when the
>>>>> new libraries change further).  You really want to track down and
>>>>> fix the root cause now, to save you hours of pain later when the
>>>>> problem recurs elsewhere.
>>>>
>>>> The thing is I don't need the libpthread.so.1 to libpthread.so.2
>>>> mapping, it works anyway (although mysql crashes) but when mapping
>>>> libpthread.so* to libthr* it works perfectly.
>>>
>>
>> The bug is well understood and is solely on the user end (operator did
>> not upgrade software correctly, and caused it to be linked to two
>> versions of the same library), it's not a software or FreeBSD bug.
>
> I don't think this is the case, if I use libthr.so it works perfectly,
> If I use libpthread.so it doesn't work. It depends on what lib I link it
> to. The mysqld was NOT linked to two versions of the same library.

Thanks for the clarification.  In that case my statement stands:

>>> It would be nice to find the bug rather than a workaround.  Can
>>> you get a stack dump of the server?

Greg
--
See complete headers for address and phone numbers.

Received on Sat Nov 05 2005 - 20:58:00 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:47 UTC