The current code assumes that encoding_needed is simply an inverse of
stream_copy, which is not true for manually attached files (for which
neither of those is true).

We already have all the necessary information in open_output_file().
This makes the information about the stream/filtergraph mappings
available earlier.

This makes the code easier to read.

This is a more appropriate place for this. H264Context.recovery_frame is
shared between frame threads, so modifying it where it is right now is
invalid.

Going through the whole decoder initialization process for a slice we
are not going to decode is unnecessary and potentially dangerous.

It is always checked in the surrounding code, so this make sure we don't
see a value from an old slice.

This is a more appropriate place for it.

This should make it more clear that this function does not need any
decoder-global state other than the parameter sets.

The code does some weird casting to a 2-dimensional sub-array of
ref2frm. This is not necessary, since only one dimension is needed
there.

Those are unused remnants of the old SVQ3 code.

This will prevent conflicts e.g. in code that deals with both h264 and
hevc.

The code does not depend on the h264 decoder anymore and only needs
information from h264_ps

This field (which the spec calls max_num_ref_frames) must be less than
or equal to MaxDpbFrames, which is at most 16.

The PS parsing code is independent from the decoder, so it makes more
sense for it to have its own separate header.

The SVQ3 decoder has been decoupled from the H.264 decoder, so it can
now use its own data type.

Move the NAL unit types into it. This will allow to stop including the
whole decoder-specific h264dec.h in some code that is unrelated to the
decoder and only needs some enum values.

This is more consistent with the naming of other decoders.

Right now this code is mixed with selecting the next output frame. Move
it to a separate function called from h264_field_start(), which is a
more appropriate place for this.

This is a more appropriate place for it.

Doing this after ff_thread_finish_setup() is called is invalid and can
conflict with reads from the other thread.

While the value of those variables will be constant for the whole frame,
they are only used in two functions called from slice header decoding.
Moving them to the per-slice context allows us to make the H264Context
passed to slice_header_parse() constant.

Copy them into the decoder-global context in field_start(). This avoids
modifying the decoder-global context during bitstream parsing.

Avoid unnecessary modification of the decoder-global state in per-slice
code.

There is no bitstream parsing in that block and messing with
decoder-global state is not something that belongs into header parsing.

Nothing else in this function depends on the value of current_slice,
except for two validity checks. Those checks are also moved out of
slice_header_parse().

Replace the decoder-global nal_unit_type/nal_ref_idc variables with the
per-NAL ones. The decoder-global ones still cannot be removed because
they are used by hwaccels.

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

Signed-off-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

Transparency is supported only by YUV and within specific bit depths.

The tests are inspired by similar tests for vp9 by
Ronald Bultje.
Signed-off-by: Martin Storsjö <martin@martin.st>

This function needs to return false, or data in the additional tables
will be skipped, and the decoder will not be able to decode frames
associated with them.

Store data from each stsd in a separate extradata buffer, keep track of
the stsc index for read and seek operations, switch buffers when the
index differs. Decoder is notified with an AV_PKT_DATA_NEW_EXTRADATA
packet side data.

Since H264 supports this notification, and can be reset midstream, enable
this feature only for multiple avcC's. All other stsd types (such as
hvc1 and hev1) need decoder-side changes, so they are left disabled for
now.

This is implemented only in non-fragmented MOVs.
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>