Martin Storsjö [Sun, 15 Mar 2015 21:40:06 +0000 (23:40 +0200)]
movenc: Only adjust the cts offset at the start of fragments if necessary
Adjusting it is only necessary when a sidx/tfrf/tfxd atom already has
been written for the previous fragment (since the sidx/tfrf/tfxd atoms
include the duration between the first pts of the previous fragment, to
the first pts of the new fragment).
Martin Storsjö [Mon, 9 Mar 2015 09:47:54 +0000 (11:47 +0200)]
movenc: Set the last packet duration based on the next packet when autoflushing
When automatically flushing fragments based on set conditions
(fragmentation on keyframes, after some interval or byte size),
we already have the next packet for one stream - use this for setting
the duration of the last packet in the flushed fragment correctly.
This avoids having to adjust the timestamp of the first packet in
the new fragment since the last duration was unknown.
Unfortunately, this only works for automatic flushing (not for
caller-triggered flushing, like in the dash muxer), and only for the
one single track that triggered the flushing. The duration of the
last sample in all other tracks still is dependent on AVPacket
duration (or heuristics).
Martin Storsjö [Fri, 6 Mar 2015 20:30:31 +0000 (22:30 +0200)]
dashenc: Heuristically fill in the duration of packets that need it
This avoids that the mp4 muxer does a similar heuristic, adjusting
the timestamps in a way that the dash muxer doesn't know the actual
timestamps written to the file in the end. By making sure that the
mp4 muxer internal heuristic isn't applied, we know the exact
timestamps written to file, so that the timestamps in manifest match
the files.
Martin Storsjö [Fri, 6 Mar 2015 13:45:50 +0000 (15:45 +0200)]
movenc: Heuristically set the duration of the last sample in a fragment if not set
Even if this is a guess, it is way better than writing a zero duration
of the last sample in a fragment (because if the duration is zero,
the first sample of the next fragment will have the same timestamp
as the last sample in the previous one).
Since we normally don't require libavformat muxer users to set
the duration field in AVPacket, we probably can't strictly require
it here either, so don't log this as a strict warning, only as info.
Martin Storsjö [Thu, 19 Mar 2015 07:28:25 +0000 (09:28 +0200)]
udp: Fix crashes after adding AVOptions
Add a missing AVClass member, check whether localaddr is null.
(Previously, localaddr was always a local stack buffer, while it
now also can be an avoption string which can be null.)
This fixes crashes when not passing any localaddr parameter, since 66028b7ba.
Derek Buitenhuis [Mon, 16 Mar 2015 20:08:35 +0000 (20:08 +0000)]
mov: Add option to keep exact packet sequence after seeking
The current behavior may produce a different sequence of packets
after seeking, compared to demuxing linearly from the beginning.
This is because the MOV demuxer seeks in each stream individually,
based on timestamp, which may set each stream at a slightly different
position than if the file would have been read sequentially.
This makes implementing certain operations, such as segmenting,
quite hard, and slower than need be.
Therefore, add an option which retains the same packet sequence
after seeking, as when a file is demuxed linearly.
Martin Storsjö [Tue, 10 Sep 2013 12:24:24 +0000 (15:24 +0300)]
movenc: Support writing multiple trun atoms
This is needed if all the data for one track isn't continuous
within the mdat. Normally we make sure all the data for one
track is continuous, but in new cases we will need to have
the samples interleaved.
Vittorio Giovara [Wed, 11 Mar 2015 20:43:59 +0000 (20:43 +0000)]
mjpeg: Mark decoder family as thread safe
No global variables are used and the VLC tables are allocated without
static elements. This will allow using a JPEG decoding context within
other decoders.
Vittorio Giovara [Wed, 11 Mar 2015 20:43:58 +0000 (20:43 +0000)]
lavc: Introduce AVCodec internal capabilities
This field is designed for marking codec properties useful to lavc internals.
Two internal capabilities are added:
- FF_CODEC_CAP_INIT_THREADSAFE: codec can be opened without locks;
- FF_CODEC_CAP_INIT_CLEANUP: codec frees memory if initialization fails.
So far it is only set in roq_encode_frame, but it is used in
roq_encode_end to free the coded_frame. This currently segfaults if
roq_encode_frame is not called between roq_encode_init and
roq_encode_end.
CC:libav-stable@libav.org Signed-off-by: Andreas Cadhalpun <Andreas.Cadhalpun@googlemail.com> Signed-off-by: Anton Khirnov <anton@khirnov.net>
rtpdec_vp9: Drop extra sanity check for size of input packet
In this case len is always at least 1, since it is checked against
RTP_VP9_DESC_REQUIRED_SIZE + 1 and then it is reduced by
RTP_VP9_DESC_REQUIRED_SIZE before entering the has_pic_id check.
Martin Storsjö [Tue, 20 Jan 2015 14:33:12 +0000 (15:33 +0100)]
movenc: Support setting fragment_index before the moov atom is written
This way, the caller doesn't need to coordinate setting the option
after the moov atom has been written. The downside is that it is
no longer possible to use the option for checking whether the moov
atom already has been written, but a caller is able to keep track
of that by other means anyway.
Martin Storsjö [Tue, 20 Jan 2015 14:54:03 +0000 (15:54 +0100)]
movenc: Don't rely on the fragment index for vc1 info gathering
The previous use of the mov->fragments field, for determining whether
written packets were part of the first fragment or not, didn't
work as intended when using the empty_moov flag.
Martin Storsjö [Mon, 9 Mar 2015 21:14:19 +0000 (23:14 +0200)]
rtpenc_mpegts: Set chain->rtp_ctx only after avformat_write_header succeeded
By making sure we at each time only have one pointer set, either a
local variable or one in the context, we avoid potential double frees
in the cleanup routines. If chain->rtp_ctx is set, it is closed by
calling avformat_write_trailer, but that shouldn't be called unless
avformat_write_header succeeded.