On Fri, Dec 16, 2016 at 05:13:50PM +0100, wm4 wrote: > On Thu, 15 Dec 2016 22:01:46 +0000 > Franklin Phillips <franklinphill...@gmx.com> wrote: > > > Hi, > > > > I tested this code by converting HLS streams into mp4 files and it > > seemed to work fine. However I recently compiled mpv with these changes > > and when I play back the Apple example stream given in ticket #2833, it > > seems to skip a lot of subtitles. It works fine for the real world > > streams I'm using it for but it's probably worth holding off with this > > change until that's worked out. > > Could be a mpv-specific issue, maybe report it there. I think you might be right, I could not reproduce it with ffmpeg alone but I also cannot report it until these changes have been merged. > > > Also I tried to make minimal changes to the existing code path because > > there don't seem to be any tests for HLS making it hard to know when > > something is broken which I guess is why there is the duplication. > > However I think you are right and it should be refactored, I had a > > further look at how it could be done and will continue working on it. > > I wonder why not. Couldn't we put a small HLS stream on disk for FATE > and let libavformat demux it? This still wouldn't test HTTP interaction > of course. This sounds like a good idea.
I have refactored the code to reduce duplication of code and sent the new patches. _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
