[FFmpeg-cvslog] r19173 - trunk/libavformat/adtsenc.c

[Reimar Döffinger]

On Tue, Jun 30, 2009 at 06:53:43PM +0100, M?ns Rullg?rd wrote:
> Libavformat should return the packets as they are in the bitstream.
> Parsing should be left to the caller, should it need it.

Not that this has anything to do with the original discussion, but
I'd guess this can be done quite simply by adding a "max parse level"
or something like that to libavformat, though then the same would
have to be used for all codecs, which might not be desired.
It is also possible to change need_parsing for each stream, but that
probably needs some changes so that libavformat does not already fail
before an application can modify it.
Not parsing by default seems like a bad idea to me, it enables the
really rare use case where someone wants to use libavformat but doesn't
have the space for libavcodec with only the parsers enabled, but increases
complexity for more common cases: decoding with libavcodec or remuxing to
a different format. For the latter case, it also introduces a risk that
they will create invalid files without noticing because e.g. a lot of H.264,
dts etc. codecs always do parsing.