On Tue, Jun 24, 2008 at 05:21:51PM +0300, Valery Masiutsin wrote: > > > > ports-lib-depends-check runs fine, however regression tests > > > > fail for swfdec at the moment (gstreamer with external ffmpeg > > > > issues). > > > > > > Such as? > > > > > > yes, please elaborate. I'm working on an ffmpeg update and would > > love to know potential issues to look for. > > Linking of external ffmpeg is the issue. By its nature, gstreamer > ffmpeg module has to have, reasonable defaults (bitrate, etc) for > different codecs, available in ffmpeg. > However codecs identifiers, changing in ffmpeg all the time, also new > codecs are being added, that was the reason why gstreamer folks > imported ffmpeg into the module itself. When you are linking with > external ffmpeg, you could easily hit this codec identifier mismatch. > As result you will often get errors - like > "Unknown codec ID 86022, please add here", and such. Just run make > regress in swfdec and you will see it.
86022 is CODC_ID_VORBIS. it exists in the ffmpeg version that ships with gst-ffmpeg-0.10.4 as well as our version in ports. in other words, it was there for months before the last release of gst-ffmpeg. the error is because CODEC_ID_VORBIS is not handled in a switch(). the error is on the part of the gst-ffmpeg developers' sloppiness (they use CODEC_ID_VORBIS in other places), not because of using an outside version of ffmpeg. otoh, gst-ffmpeg does go ahead and "work around" a "bug" in the ffmpeg that they ship, which causes problems with other versions of ffmpeg that don't have the "bug". > I am not an expert in ffmpeg and gstreamer, but this is how situation > is seen to me. > I am going to switch off gstreamer support in swfdec and turn back to > linking with ffmpeg. > > Regards Valery. > -- [EMAIL PROTECTED] SDF Public Access UNIX System - http://sdf.lonestar.org
