Apologies, I didn't realize that matroska->done was only checked in one place. But Nicolas raises a good point about the subtle change in behavior. I did not intend to change the overall behavior of the code for clients, so I'll modify the patch to restore the matroska->done check.
On Tue, Aug 9, 2016 at 5:58 AM, Nicolas George <geo...@nsup.org> wrote: > Le tridi 23 thermidor, an CCXXIV, Michael Niedermayer a écrit : > > prior to this the "done" field was used to stop further work when > > EOF was reached, now this is removed. > > The patches commit message doesnt say anything why all reads for the > > done field are removed or what if any performance impact that has > > > > the patch even adds code setting the done variable even though all > > reads of the varable are removed > > This doesnt look right > > If I read the code correctly, on the first round the result will be roughly > the same, with just a more accurate error code. The difference about the > done flag will appear if the application tries again to read a packet after > an error: without the check on done, it will try to resync again, which may > be a good thing in some cases. > > But of course, changing it inadvertently is not a good thing, and neither > keeping dead code for setting an unused flag. > > Regards, > > -- > Nicolas George > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > http://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
