> On 08/20/2010 03:59 AM, Gary R. Schmidt wrote:
>> On Thu, August 19, 2010 17:46, Bruno Friedmann wrote:
>> [SNIP]
>>>
>>> It seems we have to manage that in the configure script, which actually
>>> fail in this task.
>> Yes, I have a hand-munged configure script for Solaris 10 - static
>> libraries are deprecated and not supplied anymore, and the Webstack
>> implementation of MySQL does not supply static libraries.
>>
>>> 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.
>> Now that's amusing - if there are no symbols visible in the shared object,
>> how can anything link to it?
>>
>> Strange - I've just done a quick check across a bunch of *NIX systems, and
>> none of them have shared object without symbols, "nm x" returns great
>> gouts of verbiage on Solaris, HP-UX, AIX, Centos 5 and RHEL5 - so just
>> what OS are you on where libmysqlclient_r.so has no symbols?
>>
>> Cheers,
>> Gary B-)
>>
>>
>>
>
> Hi Gary all test made are against openSUSE version from 10.3 to 11.3 actually.
> It seems that most binaries are stripped down.
>
> but even with the -debuginfo package installed I can't obtain nm info on .so
> Some of them have info, some not.
>
> For example in 10.3 version there's a libpq.so & libpg.a
> nm libpq.a give the expected result, libpg.so = no symbols
>
> In 11.3 there's no more libpg.a but nm libpg.so give the result.
>
> Now on 10.3 or 11.3 there's no symbol for libmysqlclient_r.so
> But if you check with nm -D all symbols appears ....
>
> Sorry I'm not C programmer, and didn't understand all implications it drives.
> At the beginning, I would just understand why I can't have batch-enabled
> building bacula 5.0.3 on openSUSE,
> and where to place the libbacsql.so lib in rpm :-)
>
> I can give all type of informations needed to help finding a way.
> Just forward me a todo list.
>
> Kern did you think it's time to open a bug, with a message compilation of this
> thread ?
> (Just the part of batch-enable detection)
If I can add another data point into this... I'm running configure for
bacula 5.0.3 on RedHat Beta Enterprise Linux 6 beta refresh 2. The thread
safe detection fails for my configure too.
I have the following mysql packages installed:
# rpm -qa |grep mysql
mysql-connector-java-5.1.12-2.el6.x86_64
mysql-libs-5.1.47-3.el6.x86_64
mysql-devel-5.1.47-3.el6.x86_64
mysql-server-5.1.47-3.el6.x86_64
mysql-5.1.47-3.el6.x86_64
Here's some more info:
# find / -name libmysqlclient_r\*
/usr/lib64/mysql/libmysqlclient_r.so
/usr/lib64/mysql/libmysqlclient_r.so.16.0.0
/usr/lib64/mysql/libmysqlclient_r.so.16
# find / -name libmysqlclient_r.a
# nm /usr/lib64/mysql/libmysqlclient_r.so
nm: /usr/lib64/mysql/libmysqlclient_r.so: no symbols
# nm /usr/lib64/mysql/libmysqlclient_r.so.16
nm: /usr/lib64/mysql/libmysqlclient_r.so.16: no symbols
however as suggested above, doing nm -D produces a different result
# nm -D /usr/lib64/mysql/libmysqlclient_r.so |grep pthread_mutex_lock
U pthread_mutex_lock
Would I be safe if I just change my configure script to pass this test?
Could the detection be changed somehow to make it more robust?
Bob
------------------------------------------------------------------------------
Start uncovering the many advantages of virtual appliances
and start using them to simplify application deployment and
accelerate your shift to cloud computing.
http://p.sf.net/sfu/novell-sfdev2dev
_______________________________________________
Bacula-devel mailing list
Bacula-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-devel
