In the recorder, if the video or audio stream is disabled, do not wait
for its initialization (it will never happen) to process the header.
In that case (scrcpy --no-audio --record=file.mp4), this caused the
whole content to be buffered in memory, and written only on exit.
The sc_cond_wait() in sc_recorder_process_header() needs to be notified
of changes to video_init/audio_init (protected by stream_cond) and
video_queue/audio_queue (protected by queue_cond).
Use only one condition variable to simplify.
On initial connection, scrcpy sent some device metadata:
- the device name (to be used as window title)
- the initial video size (before any frame or even SPS/PPS)
But it is better to provide the initial video size as part as the video
stream, so that it can be demuxed and exposed via AVCodecContext to
sinks.
This avoids to pass an explicit "initial frame size" for the screen, the
recorder and the v4l2 sink.
Previously, the packet sink push() implementation just set the codec and
notified a wait condition. Then the recorder thread read the codec and
created the AVStream.
But this was racy: an AVFrame could be pushed before the creation of the
AVStream, causing its video_stream_index or audio_stream_index to be
initialized to -1.
Also, in the future, the AVStream initialization might need data
provided by the packet sink open(), so initialize it there (with a
mutex).
The packets queued for recording were wrapped in a dynamically allocated
structure with a "next" field.
To avoid this additional layer of allocation and indirection, use a
VecDeque.
By default, audio is enabled (--no-audio must be explicitly passed to
disable it).
However, some devices may not support audio capture (typically devices
below Android 11, or Android 11 when the shell application is not
foreground on start).
In that case, make the server notify the client to dynamically disable
audio forwarding so that it does not wait indefinitely for an audio
stream.
Also disable audio on unknown codec or missing decoder on the
client-side, for the same reasons.
PR #3757 <https://github.com/Genymobile/scrcpy/pull/3757>
The recorder opened the target file from the packet sink open()
callback, called by the demuxer. Only then the recorder thread was
started.
One golden rule for the recorder is to never block the demuxer for I/O,
because it would impact mirroring. This rule is respected on recording
packets, but not for the initial recorder opening.
Therefore, start the recorder thread from sc_recorder_init(), open the
file immediately from the recorder thread, then make it wait for the
stream to start (on packet sink open()).
Now that the recorder can report errors directly (rather than making the
demuxer call fail), it is possible to report file opening error even
before the packet sink is open.
The recorder has two initialization phases: one to initialize the
concrete recorder object, and one to open its packet_sink trait.
Initialize mutex and condvar as part of the object initialization.
If there were several packet_sink traits (spoiler: one for video, one
for audio), then the mutex and condvar would still be initialized only
once.
Stop scrcpy on recorder errors.
It was previously indirectly stopped by the demuxer, which failed to
push packets to a recorder in error. Report it directly instead:
- it avoids to wait for the next demuxer call;
- it will allow to open the target file from a separate thread and stop
immediately on any I/O error.
The PTS received from MediaCodec are expressed relative to an arbitrary
clock origin. We consider the PTS of the first frame to be 0, and the
PTS of every other frame is relative to this first PTS (note that the
PTS is only used for recording, it is ignored for mirroring).
For simplicity, this relative PTS was computed on the server-side.
To prepare support for multiple stream (video and audio), send the
packet with its original PTS, and handle the PTS offset on the
client-side (by the recorder).
Since we can't know in advance which stream will produce the first
packet with the lowest PTS (a packet received later on one stream may
have a PTS lower than a packet received earlier on another stream),
computing the PTS on the server-side would require unnecessary waiting.
Prefix the name of threads by "scrcpy-". This improves readability in
the output of `top -H` for example.
Limit the thread names to 16 bytes, because it is limited on some
platforms.
From FFmpeg/doc/APIchanges:
2021-03-17 - f7db77bd87 - lavc 58.133.100 - codec.h
Deprecated av_init_packet(). Once removed, sizeof(AVPacket) will
no longer be a part of the public ABI.
Refs #2302 <https://github.com/Genymobile/scrcpy/issues/2302>
The fact that the recorder uses a separate thread is an internal detail,
so the functions _start(), _stop() and _join() should not be exposed.
Instead, start the thread on _open() and _stop()+_join() on close().
This paves the way to expose the recorder as a packet sink trait.
