On Sun, 18 Aug 2013, Luca Barbato wrote:
On 18/08/13 16:19, Martin Storsjö wrote:
This looks good. What about adding an else case that returns an
error for all other unknown/unhandled AVMEDIA types? Or alternatively
just set the timescale based on time_base for them as well?
AVMEDIA_TYPE_ATTACHMENT rightfully does not have a timebase.
Hmm, indeed. But for those ones we shouldn't even create a track in the
mov, so that doesn't really apply...
AVMEDIA_TYPE_UNKNOWN is purely opaque so could have one.
If we do either of these, patch 2/3 shouldn't be necessary, right?
(Otherwise if we go for doing the check as in 2/3, we might need to
add a number of other checks around the muxer as well.)
I do not have enough experience about mov and this set is made to cope
with a crash mostly, I'll ask you on irc for mov specification so I can
see how wrong I am regarding my assumptions on timecode.
Well I don't know about timecode specifically, but in general, for tracks
that actually contain sampled data, leaving timescale at zero doesn't
sound like a good thing to do.
But without going too much deeper into these matters, applying both 2+3
for now would probably be an acceptable compromise - patch 2 isn't harmful
anyway, and then we don't need to decide on how to handle all the
other/unknown cases, yet...
// Martin
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
