On Tue, 21 Jul 2009, Julien BLACHE wrote:
Peter Denison <bug-repo...@marshadder.org> wrote:
I'm afraid you'll have to get to the bottom of this, because I don't
have a similar setup available at this time to debug this. I'll try
work on this, but real life may soon get in the way.
No problem - I will have a go. Debugging a library called from a
plugin is certainly going to sharpen my gdb skills!
Find out what buffer ebx represents, might actually not be the one I
thought it'd be. Could be the input buffer instead of the output one.
Wish I could give it a try, but anything i386 here either runs stable
or has been replaced with amd64.
Done a bit more work on this (though still haven't managed to work out how
to debug properly!). Your patch now forces it to fail, as the malloc()
will always return an address which is 8-byte-aligned, but not 16-byte
aligned, due to the arena header.
The ebx value is 0x1000 greater than the value of handle->buffer, that is
assigned from the return from malloc(). It doesn't fail on the first call
to avcodec_decode_audio2() from _ssc_ffmpeg_read_frame(), but on the
second, repeatably.
--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
