On 27/05/15 23:44, Andreas Cadhalpun wrote: > On 26.05.2015 20:48, Luca Barbato wrote: >> On 26/05/15 20:42, Andreas Cadhalpun wrote: >>> The function doesn't return error codes, which is a bit strange. >> >> Exactly that > > I think errors in mov_build_index are ignored on purpose, so adding > error codes wouldn't help a lot, because forwarding them would break > FATE tests. > >> and I'm not sure if some user would want to get this stream >> even if is "absurd", thus why I'd rather either provide the packets >> marking them as corrupted or provide a switch by using our "explode" >> feature to drop, take or stop once you find them. >> >> Would be possible to have metadata-only empty packets this way? > > As I said I think the spec implies that such a stream is invalid, > so I'd like to see a sample where it is useful, before worrying about > supporting it.
The approach is sound. > Additionally your suggested changes would defeat the primary purpose > of this patch, which is to significantly speed up processing of some > broken files. Fine by me then. lu _______________________________________________ libav-devel mailing list libav-devel@libav.org https://lists.libav.org/mailman/listinfo/libav-devel