On 08/18/2010 09:46 PM, Martin Simmons wrote:
>>>>>> On Wed, 18 Aug 2010 06:58:26 +0200, Bruno Friedmann said:
>>
>> On 08/17/2010 01:37 PM, Bruno Friedmann wrote:
>>> On 08/17/2010 01:06 PM, Martin Simmons wrote:
>>>>>>>>> On Tue, 17 Aug 2010 07:45:29 +0200, Bruno Friedmann said:
>>>>>
>>>>> On 08/16/2010 08:58 PM, Martin Simmons wrote:
>>>>>
>>>>> ok so there's a trouble that I don't understand for the moment ...
>>>>>
>>>>> I've the two lines
>>>>> checking for PQisthreadsafe in -lpq... yes
>>>>> checking for PQputCopyData in -lpq... yes
>>>>>
>>>>> also in pg_config the enable-thread-safety is here.
>>>>>
>>>>> ...
>>>>>
>>>>> Any ideas where to look ? I've also tested against the git branches with
>>>>> the same result.
>>>>
>>>> You'll have to look in the configure script to see what it does. It can
>>>> help
>>>> to run it with tracing like this:
>>>>
>>>> sh -x configure --enable-batch-insert --with-postgresql ... > config.std
>>>> 2>&1
>>>>
>>>> and then look in config.std to see exactly what happens.
>>>>
>>>> __Martin
>>>
>>> Thank Martin for the tips.
>>>
>>> I don't understand what happen here for the moment.
>>> I've attach the configure result. If you have any idea ...
>>>
>>>
>>>
>>
>> I've put the result online, as the list strip the attachment
>> http://linux.ioda.net/mirror/bacula/config.std.gz
>
> The problem is that the configure script only allows batch insert if it can
> find multithreading in libpq. It does that by
>
> test -f $SQL_LIB && nm $SQL_LIB | grep pthread_mutex_lock
>
> where $SQL_LIB is /usr/lib/libpq.a in your case. In your log, it can't find
> /usr/lib/libpq.a, so batch insert is off.
>
> I think the configure test might be broken, because it relies on the static
> library libpq.a, which often isn't installed now (e.g. see
> https://fedoraproject.org/wiki/Packaging:Guidelines#Packaging_Static_Libraries).
>
> __Martin
Thanks Martin, giving a try and replacing libpq.a by libpq.so result in a
batch-insert-enable yes configure.
(Just a horrible hack, to make proof of concept .. )
It seems we have to manage that in the configure script, which actually fail in
this task.
Now I'm trying to emulate the same thing with libmysqlclient_r.so which return
no symbols and so failed to
have the batch-insert disable each time.
From what I understand if we have libmysqlclient_r == thread safe if
libmysqlclient == not thread safe.
(that's what ldd say about the two ... )
But what about the "distant mysql server" as dir can be installed in another
place that the db server.
Is it sufficient to only have the client being thread safe ?
How would react batch-insert if the client is thread safe, but talk to a
distant server which is not ?
--
Bruno Friedmann br...@ioda-net.ch
Ioda-Net Sàrl www.ioda-net.ch
openSUSE Member
User www.ioda.net/r/osu
Blog www.ioda.net/r/blog
fsfe fellowship www.fsfe.org
(bruno.friedmann (at) fsfe.org )
tigerfoot on irc
GPG KEY : D5C9B751C4653227
------------------------------------------------------------------------------
This SF.net email is sponsored by
Make an app they can't live without
Enter the BlackBerry Developer Challenge
http://p.sf.net/sfu/RIM-dev2dev
_______________________________________________
Bacula-devel mailing list
Bacula-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-devel
