Hello Luca, I won't say thanks to snow either, since it ruined completely my Christmas Gifts marathons :)
Those are great news, thanks for your support ! Please tell me when you're about to commit, I'll grab the latest svn trunk, and test my code with it (hoping the best!) By then, I'll apply your patch and see if ffplay is no longer broken. I'll report back to you this day. Thanks a lot again, Kind regards, Guillaume. 2009/12/21 Luca Abeni <[email protected]> > Luca Abeni wrote: > [...] > > Sorry, the only thing I can suggest is a bisection: work with svn, >> starting >> from a working version, then find a non-working version, and use a >> dichotomic >> search to find the first non-working version (so, you can highlight the >> commit >> which broke this functionality). >> >> Once the particular commit that created the problem is known, it will be >> possible to understand where is the bug (maybe such a commit just exposed >> a previously existing bug). >> > > Ok, thanks to some snow and to the idiocy of the italian transportation > system, > I had plenty of spare time in this weekend... :( > > The good side of this is that I had time to bisect the regression. > Short version of the story: the regression was introduced by r19302, but > such a commit is correct and just made a different bug (in rtpdec.c) > visible... The attached patch fixes it for me. > > > The long story: rtpdec.c has been broken since forever when receiving MP3 > audio (well, technically speaking MP3 over RTP was not supported by > rtpdec.c ;-). > Basically, it did not remove the payload header from the RTP payload before > passing it to the MP3 parser. But the audio mpeg decoder in libavcodec > was pretty forgiving, silently discarding the payload header (so, the > decoder decoded broken streams without complaining). > r19302 changed this behaviour, so the decoder is now generating errors > when broken streams are provided (and this is the correct behaviour, IMHO). > The attached patch fixes rtpdec.c to remove the payload header from MP3 > streams, so that the audio decoder can correctly decode them. > This fixes the problem for me. > > If noone is against it, I'll commit this patch in few days. > > > Luca > > _______________________________________________ > libav-user mailing list > [email protected] > https://lists.mplayerhq.hu/mailman/listinfo/libav-user > > _______________________________________________ libav-user mailing list [email protected] https://lists.mplayerhq.hu/mailman/listinfo/libav-user
