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).
This has multiple advantages:
* The macro has multiple parameters that often have similar or identical
values, yet very different meanings (one is the name of the
OptionsContext member where the parsed options are stored, the other
the name of the variable into which the result is written); this
change makes each of these explicit.
* The macro returns on failure, which may cause leaks - this was the
reason for adding MATCH_PER_STREAM_OPT_CLEAN(), also ost_add()
currently leaks encoder_opts. The new function returns failure to its
caller, which decides how to deal with it. While that adds a lot of
error checks/forwards for now, those will be reduced in following
commits.
* new code is type- and const- correct
Invocations of MATCH_PER_STREAM_OPT() with other types will be converted
in following commits.
Share the code between encoding and decoding. Instead of checking every
stream's options dictionary (which is also used for other purposes),
track all used options in a dedicated dictionary.
The decision whether -apad actually does anything is made based on muxer
properties, and so more properly belongs there. Filtering code only
receives the result.
Instead pass the encoder through a newly-added output options struct,
analogous to previously added input options.
Will allow decoupling filtering from encoding in future commits.
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
The first of these binds inputs of complex filtergraphs to demuxer
streams (with a misleading comment claiming it *creates* complex
filtergraphs).
The second ensures that all filtergraph outputs are connected to an
encoder.
Merge them into a single function, which simplifies the ffmpeg_filter
API, is shorter, and will also be useful in following commits.
Also, rename misleadingly-named init_input_filter() to
fg_complex_bind_input().
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.
Do not construct the name manually from input file/stream indices.
This is a step towards avoiding the assumption that filtergraph inputs
are always fed by demuxers.
The computation is based on demuxer properties, so that is the more
appropriate place for it. Filter code just receives the desired
start time/duration.
The filename is freed with the OptionsContext and therefore
there will be a use-after-free when reporting the filename
in print_stream_maps(). So create a copy of the string.
This is a regression since 8aed3911fc.
fate-lavf-mkv_attachment exhibits it (and reports a random nonsense
filename here), but this does not make the test fail (not even with
valgrind; only with ASAN, as it aborts on use-after-free).
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
