Re: [AD] packaging trouble

[ Thread Index | Date Index | More lists.liballeg.org/allegro-developers Archives ]


#include <hallo.h>
George Foot wrote on Sat Jul 14, 2001 um 02:11:06PM:

> > objdump --headers --private-headers -T lib/unix/liballeg-3.9.37.so |grep rel.text
> >   5 .rel.text     00000c08  00010770  00010770  00010770  2**2
> > 00010770 l    d  .rel.text      00000000              
> 
> On my somewhat old 3.9.37 WIP, I get a similar result --
> slightly smaller size though.
> 
> However, I've dumped the --full-contents too, and looking in my
> editor (and guessing the format of the .rel.text chunk) I see:
> 
>     Contents of section .rel.text:
>      1239c 5d660a00 02130600 71660a00 02140600  ]f......qf......

Thanks for debugging. I have no experience in debugging low level stuff,
you showed me a way to go.

> That is two entries, in little-endian.  The first entry is:
> 
>     000a665d (address of relocation) => 00061302 (dunno what this is)
> 
> The address lies within XF86DGAQueryExtension.  (The second

It was really a Xfree problem, I reported to the maintainer about the
issue. He knows about the problem with static libs and expects a fix
from upstream soon.

> address seems to be in _mixer_exit.)  Looking on down the

_mixer_exit belongs to esd plugin - it was affected by the same problem.
Having disabled DGA, DGA2, VIDMODE and esddigi, I can not detect any
non-PIC code in the library or in the plugins. Thanks for the help! It
is not a nice solution to disable all drivers, but better than having
problems with them on the other hand.

> One thing to try is disabling DGA, and seeing if it removes the
> rel.text section (I'll do this shortly).

It works! No non-PIC code any longer. On the weekend, I received a new
bug report, complaining about linking problems with non-PIC code. In
fact, gcc on HP-PA refuses to link PIC and non-PIC objects together.

Now, there is still one problem:
misscompilation using gcc-3.0. Everything seems to compile fine (except
of tonns of error messages), but the created binaries are segfaulting
during the loading of the new library.

Tonight or tomorrow, I will try the same in another environment.
Something weird is going wrong, I have a strange feeling with this
problem.

Gruss/Regards,
Eduard.
-- 
Eine neue DAU-Generation wächst heran:
DAR = dümmster anzunehmender Administrator



Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/