GL Programs SEGV on exit()

[Peter Jeremy]

On 2012-May-08 11:46:05 +0300, Konstantin Belousov <kostikbel at gmail.com> wrote:
>On Tue, May 08, 2012 at 06:32:02AM +1000, Peter Jeremy wrote:
>> On 2012-May-07 15:13:52 +0300, Konstantin Belousov <kostikbel at gmail.com> wrote:
>> >On Mon, May 07, 2012 at 09:49:24PM +1000, Peter Jeremy wrote:
>> >> Since rebuilding WITH_NEW_XORG, I've been seeing segmentation
>> >> violations in GL programs under some conditions.  As a case that is
>> >> quick and easy to reproduce, "glxinfo -b" with software GL (which is
>> >> easily forced by running glxinfo in a X11-over-SSH session) will
>> >> core dump on exit.
>> (Due to atexit processing invoking a function in an unloaded .so)
...
>> I notice r211706 was initially marked "MFC after 3 weeks" but this was
>> never done.  I've found one posting that blames it for problems in
>> emacs but nothing else.  Is there a reason it wasn't MFC'd?
>
>It requires r211705 to get needed support from rtld. I mostly give up
>merging rtld changes to stable/8 long time ago.
>
>If you can test this all, I would merge.

I've updated my 8.x system to r235137 and merged r211705 and r211706
and neither glxinfo nor firefox now coredump.  Based on that, I'd say
r211705 and r211706 are prerequisites to running Mesa/dri 7.11.2 (part
of WITH_NEW_XORG) on 8.x (and presumably 7.x).

Thanks for the pointer.

-- 
Peter Jeremy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
Url : http://lists.freebsd.org/pipermail/freebsd-x11/attachments/20120509/3a54bd5e/attachment.pgp