Re: [Ekiga-devel-list] Ekiga 3.3.0 crash on h263 and h263-1998

[Eugen Dedu <Eugen Dedu pu-pm univ-fcomte fr>]

On 21/01/11 06:52, g trentalancia libero it wrote:
> I am not sure it is related to the bug you mentioned Eugen. It depends
> on Opal for sure. The context is not being initialised. It might depend
> on ffmpeg if, for example, avcodec_alloc_context() has failed, but I had
> no time to check it further.
>
> In any case, I have managed to create a temporary fix, which works for
> me:
>
> http://sourceforge.net/tracker/?
> func=detail&aid=3163160&group_id=204472&atid=989748

Ok, thank you.  I see that it was partially committed.  Even after this 
commit, I receive a crash as soon as the connection is established 
(using stable branch for opal and git ekiga):

#5  __libc_free (mem=0x51) at malloc.c:3710
        ar_ptr = <value optimized out>
        p = <value optimized out>
#6  0x00007f7d899c8e5d in 
H263_RFC2429_EncoderContext::~H263_RFC2429_EncoderContext (this=0x7f7d840db5
a0, __in_chrg=<value optimized out>) at h263-1998.cxx:779
No locals.
#7  0x00007f7d899c9309 in create_encoder (codec=<value optimized out>) 
at h263-1998.cxx:1351
        context = 0x7f7d840db5a0

i.e. at the line:
    av_free(_inputFrame);

Shouldn't the whole patch be committed?

Anyway, I remember having two crashes with H263, one because of free, 
and another inside ffmpeg.  The free might be fixed, but the other one 
probably not.

Cheers,
Eugen

> On Thu, 20/01/2011 at 09.04 +0100, Eugen Dedu wrote:
> > On 20/01/11 05:42, g trentalancia libero it wrote:
> > > Ekiga version 3.3.0
> > > Opal version 3.8.3
> > > Ptlib version 2.8.3
> > >
> > > ffmpeg-0.6.1
> > >
> > > This crash is very easy to reproduce and happens always. Just call any
> number
> > > using any
> > > audio codec and the H263 or H263-1998 video codec. Ekiga crashes
> > > immediately after the call begins.
> > >
> > > This does not occur when no video codec is selected or any of the other
> > > video codecs is used (theora, h261, mp4v-es).
> > >
> > > Attached stack backtrace.
> >
> > Thank you for the stack.  This is a known issue, see
> > https://bugzilla.gnome.org/show_bug.cgi?id=606265 and help if you can.