Up until now if avcodec_send_packet() returned an error,
no attempt to receive a frame from the decoder would be made.
Instead the decoder was presumed to be drained, so that more
packets could be sent to it. Yet this need not be so:
It can happen that a packet would decode to multiple frames
and that decoding the first of these (the one that is decoded
during the avcodec_send_packet() call) returns an error,
in which case the decoder is not yet ready to receive more
input as the remaining parts of the packet have not been decoded
yet. In this case, an AERROR_BUG is triggered.
This happens in ticket #11553 with VP9 (which uses a BSF
to split VP9 superframes and is therefore affected by this).
Fix this by always calling avcodec_receive_frame() unless
xerror is set.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
We can't use AV_FRAME_SIDE_DATA_FLAG_REPLACE here because the side data already in the
frame should have priority over the global one, so just ensure we don't copy any if
it already exists.
Fixes ticket #11468.
Signed-off-by: James Almer <jamrial@gmail.com>
Global side data as exported by a decoder may no longer apply if a filter in
the chain altered the frames in some form, like changing color, dimensions,
or channel layout information.
After this change, any such changes in side data will be taken into account by
the encoder futher in the process.
Signed-off-by: James Almer <jamrial@gmail.com>
This extends the syntax for specifying input streams in -map and complex
filtergraph labels, to allow selecting a view by view ID, index, or
position. The corresponding decoder is then set up to decode the
appropriate view and send frames for that view to the correct
filtergraph input(s).
There are lots of files that don't need it: The number of object
files that actually need it went down from 2011 to 884 here.
Keep it for external users in order to not cause breakages.
Also improve the other headers a bit while just at it.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
lavfi does not require aligned buffers, so we can safely apply top/left
cropping by any amount, without passing any special flags to lavc.
Longer term, an even better solution would probably be auto-inserting
the crop filter (or its hwaccel versions) as needed.
Multiple FATE tests no longer need -flags unaligned.
Do not pass an options dictionary to avcodec_open2().
This should be equivalent to current behaviour, but will allow
overriding caller-supplied options in a cleaner and more robust manner.
We can now set the COPY_OPAQUE flag directly rather going through
dec_opts.
Since e0da916b8f5b079a4865eef7f64863f50785463d the ffmpeg utility has
held multiple frames output by the decoder in internal queues without
telling the decoder that it is going to do so. When the decoder has a
fixed-size pool of frames (common in some hardware APIs where the output
frames must be stored as an array texture) this could lead to the pool
being exhausted and the decoder getting stuck. Fix this by telling the
decoder to allocate additional frames according to the queue size.
These functions used to be passed directly to pthread_create(), which
required them to return void*. This is no longer the case, so they can
return a plain int.
Current callstack looks like this:
* ifilter_bind_ist() (filter) calls ist_filter_add() (demuxer);
* ist_filter_add() opens the decoder, and then calls
dec_add_filter() (decoder);
* dec_add_filter() calls ifilter_parameters_from_dec() (i.e. back into
the filtering code) in order to give post-avcodec_open2() parameters
to the filter.
This is unnecessarily complicated. Pass the parameters as follows
instead:
* dec_init() (which opens the decoder) returns post-avcodec_open2()
parameters to its caller (i.e. the demuxer) in a parameter-only
AVFrame
* the demuxer passes these parameters to the filter in
InputFilterOptions, together with other filter options
Rename dec_open to dec_init(), as it is more descriptive of its new
purpose.
Will be useful in following commits, which will add a new path for
opening decoders.
Previously, the demuxer would register decoder with the scheduler, using
InputStream as opaque, and pass the scheduling index to the decoder.
Now the registration is done by the decoder itself, using DecoderPriv as
opaque, and the scheduling index is returned to demuxer from dec_open().
decoder_thread() then no longer needs to be accessed from outside of
ffmpeg_dec and can be made static.
It is done based on demuxer information, so that is the more appropriate
place for this code.
This is a step towards decoupling Decoder and InputStream.
* as this decision is based on demuxing information, move it from the
decoder to the demuxer
* as the issue being addressed is latency added by frame threading, we
only need to disable frame threading, not all threading
Similar to what is currently done for other components, e.g. (de)muxers.
There is nothing in the public part currently, but that will change in
future commits.
Change the main loop and every component (demuxers, decoders, filters,
encoders, muxers) to use the previously added transcode scheduler. Every
instance of every such component was already running in a separate
thread, but now they can actually run in parallel.
Changes the results of ffmpeg-fix_sub_duration_heartbeat - tested by
JEEB to be more correct and deterministic.
