Quoting Vittorio Giovara (2015-11-17 23:41:25) > On Sun, Nov 8, 2015 at 8:02 PM, Vittorio Giovara > <[email protected]> wrote: > > On Sun, Nov 8, 2015 at 1:30 PM, Anton Khirnov <[email protected]> wrote: > >> Quoting Andreas Cadhalpun (2015-11-07 17:34:15) > >>> On 07.11.2015 14:52, Luca Barbato wrote: > >>> > On 07/11/15 00:08, Andreas Cadhalpun wrote: > >>> > > >>> >> 1: > >>> >> https://sources.debian.net/src/opal/3.10.10~dfsg2-1/plugins/video/H.263-1998/h263-1998.cxx/?hl=600#L600 > >>> > > >>> > Downstream dropped it since a while. > >>> > >>> Not quite, they just moved it elsewhere: > >>> http://sourceforge.net/p/opalvoip/code/HEAD/tree/opal/trunk/plugins/video/common/ffmpeg.cxx > > > > Thanks for the analysis Andreas and Luca. > > > >> Then I guess someone should ask them if they are sure there are real > >> benefits to using that callback. Vittorio, will you do that? > > > > Sure. > > -- > > Vittorio > > A small update, I emailed the opal mailing list and the people listed > under 'commercial support' and received no reply after about 10 days. > However I have been contacted off-list by one of the developer from > H323Plus (a package which is not on Debian afaik) who mentioned that > this callback is used in his project. Unfortunately he's been pretty > vague on how the callback employed and why it is actually needed, and > argued that should it be removed noone would know how to update the > encoder. From what I can tell from the build instructions, their code > requires a patched avcodec.h (at version 0.4.7) in order to compile, > and it already employs workarounds in order to use this API (quoting > "sometimes, FFmpeg encodes the same frame multiple times").
It seems that file in h323plus is directly lifted from opal. And "nobody knows how to update the code" looks like cargo cult programming to me and is IMO not a sufficient reason to keep the field. Especially since they carry an ancient internal fork of ffmpeg. So, unless there are more substantial arguments against, I'd remove this field. -- Anton Khirnov _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
