#10078: speedhq: ffmpeg fails to decode it's own created file - low resolution
-------------------------------------+-------------------------------------
             Reporter:  ami_stuff    |                    Owner:  (none)
                 Type:  defect       |                   Status:  new
             Priority:  normal       |                Component:  avcodec
              Version:  git-master   |               Resolution:
             Keywords:  shq,         |               Blocked By:
  speedhq, speedhqenc                |
             Blocking:               |  Reproduced by developer:  1
Analyzed by developer:  0            |
-------------------------------------+-------------------------------------
Changes (by Tomas Härdin):

 * keywords:  shq => shq, speedhq, speedhqenc
 * reproduced:  0 => 1

Comment:

 I ran across this the other day with parallellizing speedhqdec, trying to
 make a non multiple of 16 width sample

 {{{
 $ ffmpeg -t 1 -f lavfi -i testsrc2 -vcodec speedhq speedhq.mov
 $ ffmpeg -t 1 -f lavfi -i testsrc2=s=328x240 -vcodec speedhq speedhq2.mov
 }}}

 Playing speedhq.mov with ffplay or mpv works fine whereas speedhq2.mov
 causes mpv to output a bunch of "Error while decoding frame!" and ffplay
 to just do nothing. This both with the current Debian stable version of
 ffmpeg (5.1.4-0+deb12u1) and with the latest master (cdcb4b9). Looking
 closer the error happens in decode_speedhq_border() as one would suspect
 in this case.

 I've also noticed that speedhqenc happily accepts non multiple of 8 input
 as well, which it probably shouldn't.

 I'm not versed in mpegvideo_enc internals, but I suspect the error is in
 there somewhere. A stop gap solution could be to only allow width to be a
 multiple of 16 in speedhqenc
-- 
Ticket URL: <https://trac.ffmpeg.org/ticket/10078#comment:3>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker
_______________________________________________
FFmpeg-trac mailing list
[email protected]
https://ffmpeg.org/mailman/listinfo/ffmpeg-trac

To unsubscribe, visit link above, or email
[email protected] with subject "unsubscribe".

Reply via email to