Closed GoogleCodeExporter closed 9 years ago
I'm wondering whether pkt.destruct() is ever anything other than
av_free(pkt.data).
If pkt.destruct() can be more than av_free(pkt.data) then perhaps
ist.pkt_data_to_free should store the pkt.destruct to call later, instead of
storing pkt.data.
FYI, I'm posting this issue in case I cannot work out a fix for myself. I may
get back later, or talk on IRC.
Original comment by mark.him...@gmail.com
on 30 Jun 2014 at 11:17
It's due to interlacing scaling with rawdecoding and uyvy422, frame is not
allocated but from the demuxer itself.
Original comment by baptiste...@gmail.com
on 1 Jul 2014 at 1:22
Ok - interesting...
As a workaround, I can fix the segfault by adding one EXTRA scale & format pair
in the command line, so instead of
-vf scale=1920:1080:interl=1,format=uyvy422
(which segfaults) I can do:
-vf
scale=1920:1080:interl=1,format=yuv422p,scale=1920:1080:interl=1,format=uyvy422
(which does not segfault)
The thought of how that 'fixes' the problem scares me...
Original comment by mark.him...@gmail.com
on 1 Jul 2014 at 1:34
It fixes it because the frame will "become" allocated by the filter chain in
the middle. I think it is just a buffer "padding" issue in that particular case.
Original comment by baptiste...@gmail.com
on 1 Jul 2014 at 7:21
Should be fixed in 0.7.2
Original comment by baptiste...@gmail.com
on 22 Dec 2014 at 3:48
Original comment by baptiste...@gmail.com
on 22 Dec 2014 at 3:48
Original issue reported on code.google.com by
mark.him...@gmail.com
on 30 Jun 2014 at 11:13