The server needs to interrupt the sockets on stop, but it must not close
them while other threads may attempt to read from or write to them.
In particular, the video_socket is read by the stream thread, and the
control_socket is written by the controller and read by receiver.
Therefore, close the socket only on sc_server_destroy(), which is called
after all other threads are joined.
Reported by TSAN on close:
WARNING: ThreadSanitizer: data race (pid=3287612)
Write of size 8 at 0x7ba000000080 by thread T1:
#0 close ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:1690 (libtsan.so.0+0x359d8)
#1 net_close ../app/src/util/net.c:280 (scrcpy+0x23643)
#2 run_server ../app/src/server.c:772 (scrcpy+0x20047)
#3 <null> <null> (libSDL2-2.0.so.0+0x905a0)
Previous read of size 8 at 0x7ba000000080 by thread T16:
#0 recv ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:6603 (libtsan.so.0+0x4f4a6)
#1 net_recv_all ../app/src/util/net.c:228 (scrcpy+0x234a9)
#2 stream_recv_packet ../app/src/stream.c:33 (scrcpy+0x2045c)
#3 run_stream ../app/src/stream.c:228 (scrcpy+0x21169)
#4 <null> <null> (libSDL2-2.0.so.0+0x905a0)
Refs ddb9396743
If --no-control is requested, then the controller instance is not
initialized. However, its reference was still passed to screen and
input_manager.
Instead, pass NULL if no controller is available.
The optimal initial size was computed from the expected dimensions, sent
immediately by the server before encoding any video frame.
However, the actual frame size may be different, for example when the
device encoder does not support the requested size.
To always handle this case properly, position and size the window only
once the first frame size is known.
PR #2947 <https://github.com/Genymobile/scrcpy/pull/2947>
Show the window only after the actual frame size is known (and if no
error has occurred).
This will allow to properly position and size the window when the size
of the first frame is different from the size initially announced by the
server.
PR #2947 <https://github.com/Genymobile/scrcpy/pull/2947>
The components should be configurable independently of the global
scrcpy_options instance: their configuration could be provided
separately, like it is the case for example for some screen parameters.
For consistency, keyboard injection should not depend on scrcpy_options.
In relative mouse mode, the mouse pointer must be "captured" from the
computer.
Toggle (disable/enable) relative mouse mode using any of the hardcoded
capture keys:
- left-Alt
- left-Super
- right-Super
These capture keys do not conflict with shortcuts, since a shortcut is
always a combination of the MOD key and some other key, while the
capture key triggers an action only if it is pressed and released alone.
The relative mouse mode is also automatically enabled on any click in
the window, and automatically disabled on focus lost (it is possible to
lose focus even without the mouse).
The decision to not send motion events when no click is pressed is
specific to Android mouse injection. Other mouse processors (e.g. for
HID mouse) will need to receive all events.
The default mouse injection works in absolute mode: it forwards clicks
at a specific position on screen.
To support HID mouse, add a flag to indicate that the mouse processor
works in relative mode: it forwards mouse motion vectors, without any
absolute reference to the screen.
A scroll event might be produced when a mouse button is pressed (for
example when scrolling while selecting a text). For consistency, pass
the actual buttons state (instead of 0).
In practice, it seems that this use case does not work properly with
Android event injection, but it will work with HID mouse.
The input_manager is strongly tied to the screen, it could not work
independently of the specific screen implementation.
To implement a user-friendly HID mouse behavior, some SDL events
will need to be handled both by the screen and by the input manager. For
example, a click must typically be handled by the input_manager so that
it is forwarded to the device, but in HID mouse mode, the first click
should be handled by the screen to capture the mouse (enable relative
mouse mode).
Make the input_manager a descendant of the screen, so that the screen
decides what to do on SDL events.
Concretely, replace this structure hierarchy:
+- struct scrcpy
+- struct input_manager
+- struct screen
by this one:
+- struct scrcpy
+- struct screen
+- struct input_manager
This avoids to directly pass the options instance (which contains more
data than strictly necessary), and limit the number of parameters for
the init function.
Not all key processors support text injection (HID keyboard does not
support it).
Instead of providing a dummy op function, set it to NULL and check on
the caller side before calling it.
Pass scrcpy input events instead of SDL input events to mouse
processors.
These events represent exactly what mouse processors need, abstracted
from any visual orientation and scaling applied on the SDL window.
This makes the mouse processors independent of the "screen" instance,
and the implementation source code independent of the SDL API.
This aims to make the key/mouse processors independent of the "screen",
by processing scrcpy-specific input events instead of SDL events.
In particular, these scrcpy events are not impacted by any UI window
scaling or rotation (contrary to SDL events).
The input manager exposed functions taking an "actions" parameter,
containing a bitmask-OR of ACTION_UP and ACTION_DOWN.
But they are never called with both actions simultaneously anymore, so
simplify.
Refs 964b6d2243
Refs d0739911a3
This allows to report a meaningful error message if an unsupported
feature is used on an incompatible platform. This is consistent with the
behavior of -K/--hid-keyboard.
The "resize to fit" feature (MOD+w or double-click on black borders)
computed the "optimal size" using the same function computing the
initial window size on start.
However, on "resize to fit", only the black borders must be removed (the
content size must be preserved), so the display bounds must not be
considered.
Since commit 0426708544, the server is run
in a dedicated thread. For SDL, many signals, including SIGINT and
SIGTERM, are masked for new threads. As a result, if the adb server is
not already running, adb commands invoked by scrcpy will start an adb
server that ignores those signals and cannot be terminated at system
shutdown.
Fixes#2873 <https://github.com/Genymobile/scrcpy/issues/2873>
PR #2870 <https://github.com/Genymobile/scrcpy/pull/2870>
Signed-off-by: Romain Vimont <rom@rom1v.com>
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.
The sockets were never interrupted or closed by the client since recent
changes to run the server from a dedicated thread (see commit
0426708544).
As a side effect, the server could never terminate properly (it was
waiting on socket blocking calls), so it was always killed by the client
after the WATCHDOG_DELAY.
Interrupt the sockets on stop to give the servera chance to terminate
property, then close them.
If --no-control is enabled, then it is not necessary to create a second
communication socket between the client and the server.
This also facilitates the use of the server alone (without the client)
to receive only the raw video stream.
Thank you clang:
../app/src/control_msg.c:45:5: warning: suspicious concatenation of
string literals in an array initialization; did you mean to separate
the elements with a comma? [-Wstring-concatenation]
"hover-exit",
^
In function ‘memcpy’,
inlined from ‘control_msg_serialize.constprop’ at ../app/src/control_msg.c:77:5,
inlined from ‘run_controller’ at ../app/src/controller.c:69:12:
/usr/include/x86_64-linux-gnu/bits/string_fortified.h:34:10:
warning: ‘__builtin___memcpy_chk’ writing 262138 bytes into a region
of size 262130 overflows the destination [-Wstringop-overflow=]
return __builtin___memcpy_chk (__dest, __src, __len, __bos0 (__dest));
Refs 901d837165
PR #2859 <https://github.com/Genymobile/scrcpy/pull/2859>
Signed-off-by: Yu-Chen Lin <npes87184@gmail.com>
Signed-off-by: Romain Vimont <rom@rom1v.com>
There was only two key injection modes:
- the default one
- the mode with --prefer-text enabled
To prepare the addition of another mode (--raw-key-events), use an enum
instead of a bool.
PR #2831 <https://github.com/Genymobile/scrcpy/pull/2831>
Expose an option to automatically configure and reconnect the device
over TCP/IP, to simplify wireless connection without using adb
explicitly.
There are two variants:
- If a destination address is provided, then scrcpy connects to this
address before starting. The device must listen on the given TCP port
(default is 5555).
- If no destination address is provided, then scrcpy attempts to find
the IP address of the current device (typically connected over USB),
enables TCP/IP mode, then connects to this address before starting.
PR #2827 <https://github.com/Genymobile/scrcpy/pull/2827>
This will allow to read the property "service.adb.tcp.port" to know if
the TCP/IP mode is enabled on the device, and which listening port is
used.
PR #2827 <https://github.com/Genymobile/scrcpy/pull/2827>
Depending on the platform and adb versions, the lines output by adb
could end with "\r\r\n". This util function helps to remove all trailing
'\r'.
PR #2827 <https://github.com/Genymobile/scrcpy/pull/2827>
"adb connect" always returns successfully (with exit code 0), even in
case of failure.
As a workaround, capture its output and check if it starts with
"connected".
PR #2827 <https://github.com/Genymobile/scrcpy/pull/2827>
In addition to disable stdout and stderr of the child process, add a
flag to disable the error log printed by scrcpy if the command failed.
This will we useful for commands which are expected to fail in some
cases (like "adb disconnect" if the device is not connected).
PR #2827 <https://github.com/Genymobile/scrcpy/pull/2827>
Let the caller decide if stdout and stderr must be inherited on process
creation, i.e. if stdout and stderr of the child process should be
printed in the scrcpy console.
This allows to get output and errors for specific adb commands depending
on the context.
PR #2827 <https://github.com/Genymobile/scrcpy/pull/2827>
If SOCK_CLOEXEC exists, then set the flag on socket creation.
Otherwise, use fcntl() (or SetHandleInformation() on Windows) to set the
flag afterwards.
This avoids the sockets to be inherited in child processes.
Refs #2783 <https://github.com/Genymobile/scrcpy/pull/2783>
The options values to configure the server were identified by their
command-line argument index. Now that there are a lot of arguments, many
of them being booleans, it became unreadable and error-prone.
Identify the arguments by a key string instead, and make them optional.
This will also simplify running the server manually for debugging.
To allow seamless copy-paste, on Ctrl+v, a SET_CLIPBOARD request is
performed before injecting Ctrl+v.
But when HID keyboard is enabled, the Ctrl+v injection is not sent on
the same channel as the clipboard request, so they are not serialized,
and may occur in any order. If Ctrl+v happens to be injected before the
new clipboard content is set, then the old content is pasted instead,
which is incorrect.
To minimize the probability of occurrence of the wrong order, a delay of
2 milliseconds was added before injecting Ctrl+v. Then 5ms. But even
with 5ms, the wrong behavior sometimes happens.
To handle it properly, add an acknowledgement mechanism, so that Ctrl+v
is injected over AOA only after the SET_CLIPBOARD request has been
performed and acknowledged by the server.
Refs e4163321f0
Refs 45b0f8123a
PR #2814 <https://github.com/Genymobile/scrcpy/pull/2814>
Pass the information that device clipboard has been set to the key
processor. This avoids the keyprocessor to "guess", and paves the way to
implement a proper acknowledgement mechanism.
PR #2814 <https://github.com/Genymobile/scrcpy/pull/2814>
Tunnel host and port are only meaningful in "adb forward" mode.
They indicate where the client must connect to communicate with the
server. In "adb reverse" mode, the client _listens_, it does not
_connect_.
Refs #2807 <https://github.com/Genymobile/scrcpy/pull/2807>
In "adb forward" mode, by default, scrcpy connects to localhost:PORT,
where PORT is the local port passed to "adb forward". This assumes that
the tunnel is established on the local host with a local adb server
(which is the common case).
For advanced usage, add --tunnel-host and --tunnel-port to force the
connection to a different destination.
Fixes#2801 <https://github.com/Genymobile/scrcpy/issues/2801>
PR #2807 <https://github.com/Genymobile/scrcpy/pull/2807>
Signed-off-by: Romain Vimont <rom@rom1v.com>
Inconditionnally print the scrcpy version first, without using LOGI().
The log level is configured only after parsing the command line
parameters (it may be changed via -V), and command line parsing logs
should not appear before the scrcpy version.
The output of -h/--help was printed on stderr, although it is not an
error.
Printing on stdout allows to pipe the result directly:
scrcpy --help | less
Instead of (in bash):
scrcpy --help |& less
This allows to execute all adb commands with the specific -s parameter,
even if it is not provided by the user.
In practice, calling adb without -s works if there is exactly one device
connected. But some adb commands (for example "adb push" on drag & drop)
could be executed after another device is connected, so the actual
device serial must be known.
The interruptible version of the function to check process success
(sc_process_check_success_intr()) did not accept a close parameter to
avoid a race condition. But as the result, the processes were not closed
at all.
Add a close parameter, and close the process separately to avoid the
race condition.
Interrupt any blocking call on process terminated, like on
server_stop().
This allows to interrupt any blocking accept() with correct
synchronization without additional complexity.
Define server callbacks, start the server asynchronously and listen to
connection events to initialize scrcpy properly.
It will help to simplify the server code, and allows to run the UI event
loop while the server is connecting. In particular, this will allow to
receive SIGINT/Ctrl+C events during connection to interrupt immediately.
Currently, server_stop() is called from the same thread as
server_connect_to(), so interruption may never happen.
This is a step to prepare executing the server from a dedicated thread.
Use the option descriptions to generate the optstring and longopts
parameters for the getopt_long() command.
That way, the options are completely described in a single place.
On Linux, socket functions are unblocked by shutdown(), but on Windows
they are unblocked by closesocket().
Expose net_interrupt() and net_close() to abstract these differences:
- net_interrupt() calls shutdown() on Linux and closesocket() on
Windows (if not already called);
- net_close() calls close() on Linux and closesocket() on Windows (if
not already called).
This simplifies the server code, and prevents a data race on close
(reported by TSAN) on Linux (but does not fix it on Windows):
WARNING: ThreadSanitizer: data race (pid=836124)
Write of size 8 at 0x7ba0000000d0 by main thread:
#0 close ../../../../src/libsanitizer/tsan/tsan_interceptors_posix.cpp:1690 (libtsan.so.0+0x359d8)
#1 net_close ../app/src/util/net.c:211 (scrcpy+0x1c76b)
#2 close_socket ../app/src/server.c:330 (scrcpy+0x19442)
#3 server_stop ../app/src/server.c:522 (scrcpy+0x19e33)
#4 scrcpy ../app/src/scrcpy.c:532 (scrcpy+0x156fc)
#5 main ../app/src/main.c:92 (scrcpy+0x622a)
Previous read of size 8 at 0x7ba0000000d0 by thread T6:
#0 recv ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:6603 (libtsan.so.0+0x4f4a6)
#1 net_recv ../app/src/util/net.c:167 (scrcpy+0x1c5a7)
#2 run_receiver ../app/src/receiver.c:76 (scrcpy+0x12819)
#3 <null> <null> (libSDL2-2.0.so.0+0x84f40)
When Ctrl+v is pressed, a control is sent to the device to set the
device clipboard before injecting Ctrl+v.
With the InputManager method, it is guaranteed that the device
synchronization is executed before handling Ctrl+v, since the commands
are executed on the device in sequence.
However, HID are injected from the computer, so there is no such
guarantee. As a consequence, on Android, Ctrl+v triggers a paste with
the old clipboard content.
To workaround the issue, wait a bit (2 milliseconds) from the AOA
thread before injecting the event, to leave enough time for the
clipboard to be set before injecting Ctrl+v.
When an AOA HID keyboard is registered, CAPSLOCK and NUMLOCK are both
disabled, regardless of the state of the computer keyboard.
To synchronize the state, on first key event, inject CAPSLOCK and/or
NUMLOCK if necessary.
The serial is necessary to find the correct Android device for AOA.
If it is not explicitly provided by the user via -s, then execute "adb
getserialno" to retrieve it.
The AVOutputFormat name is a comma-separated list. In theory, possible
names for V4L2 are:
- "video4linux2,v4l2"
- "v4l2,video4linux2"
- "v4l2"
- "video4linux2"
To find the muxer in all cases, we must request exactly one muxer name
at a time.
PR #2718 <https://github.com/Genymobile/scrcpy/pull/2718>
Co-authored-by: Romain Vimont <rom@rom1v.com>
Signed-off-by: Romain Vimont <rom@rom1v.com>
The first frames are typically received and decoded with more delay than
the others, causing a wrong slope estimation on start.
To compensate, assume an initial slope of 1, then progressively use the
estimated slope.
To minimize latency (at the cost of jitter), scrcpy always displays a
frame as soon as it available, without waiting.
However, when recording (--record), it still writes the captured
timestamps to the output file, so that the recorded file can be played
correctly without jitter.
Some real-time use cases might benefit from adding a small latency to
compensate for jitter too. For example, few tens of seconds of latency
for live-streaming are not important, but jitter is noticeable.
Therefore, implement a buffering mechanism (disabled by default) to add
a configurable latency delay.
PR #2417 <https://github.com/Genymobile/scrcpy/issues/2417>
Currently, a frame is available to the consumer as soon as it is pushed
by the producer (which can detect if the previous frame is skipped).
Notify the new frames (and frame skipped) via callbacks instead.
This paves the way to add (optional) buffering, which will introduce a
delay between the time when the frame is produced and the time it is
available to be consumed.
The current video buffer only stores one pending frame.
In order to add a new buffering feature, move this part to a separate
"frame buffer". Keep the video_buffer, which currently delegates all its
calls to the frame_buffer.
To fix a data race, commit 5caeab5f6d
called video_buffer_push() and video_buffer_consume() under the
v4l2_sink lock.
Instead, use the previous_skipped indication (initialized with video
buffer locked) to lock only for protecting the has_frame flag.
This enables the possibility for the video_buffer to notify new frames
via callbacks without lock inversion issues.
The function sc_cond_timedwait() accepted a parameter representing the
max duration to wait, because it internally uses SDL_CondWaitTimeout().
Instead, accept a deadline, to be consistent with
pthread_cond_timedwait().
The v4l2_sink implementation directly read the internal video_buffer
field "pending_frame_consumed", which is protected by the internal
video_buffer mutex. But this mutex was not locked, so reads were racy.
Lock using the v4l2_sink mutex in addition, and use a separate field to
avoid depending on the video_buffer internal data.
Commit 21d206f360 added mutex assertions.
However, the "locker" variable to trace the locker thread id was read
and written by several threads without any protection, so it was racy.
Reported by TSAN.
The options --no-display and --no-control are independent.
The controller was not initialized when no display was requested,
because it was assumed that no control could occur without display. But
that's not true (anymore): for example, it is possible to pass
--turn-screen-off.
Fixes#2426 <https://github.com/Genymobile/scrcpy/issues/2426>
Mouse motion events were forwarded as soon as any mouse button was
pressed.
Instead, only consider left-click (and also middle-click and right-click
if --forward-all-clicks is enabled).
Change the default push target from /sdcard/ to /sdcard/Download/.
Pushing to the root of /sdcard/ is not very convenient, many apps do not
expose its content directly.
It can still be changed by --push-target.
PR #2384 <https://github.com/Genymobile/scrcpy/pull/2384>
When removing the black borders (by double-clicking on them, or by
pressing MOD+w), the window is resized to fit the device screen, but its
top-left position was left unchanged.
Instead, move the window so that the new window area is at the center of
the old window area.
Refs #2387 <https://github.com/Genymobile/scrcpy/issues/2387>
It should not be necessary, since screen_render() is called just after
on SDL_WINDOWEVENT_EXPOSED, but in practice the window content might not
be correctly displayed on restored if a rotation occurred while
minimized.
Note that calling screen_render() twice in a row on
SDL_WINDOWEVENT_EXPOSED also "fixes" the issue.
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>
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>
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>
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.
Remove the has_pending boolean, which can be replaced by:
stream->pending != NULL
Refs #2302 <https://github.com/Genymobile/scrcpy/issues/2302>
The input manager was partially initialized statically, but a call to
input_manager_init() was needed anyway, so initialize all the fields
from the "constructor".
This is consistent with the initialization of the other structs.
The frame can be unref immediately after it is pushed to the frame
sinks.
It was not really a memory leak because the frame was unref every time
by avcodec_receive_frame() (and freed on close), but a reference was
unnecessarily kept for too long.
Add a new mode to the --lock-video-orientation option, to lock the
initial orientation of the device.
This avoids to pass an explicit value (0, 1, 2 or 3) and think about
which is the right one.
The screen may not be destroyed immediately on close to avoid undefined
behavior, because it may still receive events from the decoder.
But the visual window must still be closed immediately.
The destruction order is important, but tricky, because the screen is
open/close by the decoder, but destroyed by scrcpy.c on the main thread.
Add assertions to guarantee that the screen is not destroyed before
being closed.
The video buffer is now an internal detail of the screen component.
Since the screen is plugged to the decoder via the frame sink trait, the
decoder does not access to the video buffer anymore.
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.
The video buffer took ownership of the producer frame (so that it could
swap frames quickly).
In order to support multiple sinks plugged to the decoder, the decoded
frame must not be consumed by the display video buffer.
Therefore, move the producer and consumer frames out of the video
buffer, and use FFmpeg AVFrame refcounting to share ownership while
avoiding copies.
This flag forced the decoder to wait for the previous frame to be
consumed by the display.
It was initially implemented as a compilation flag for testing, not
intended to be exposed at runtime. But to remove ifdefs and to allow
users to test this flag easily, it had finally been exposed by commit
ebccb9f6cc.
In practice, it turned out to be useless: it had no practical impact,
and it did not solve or mitigate any performance issues causing frame
skipping.
But that added some complexity to the codebase: it required an
additional condition variable, and made video buffer calls possibly
blocking, which in turn required code to interrupt it on exit.
To prepare support for multiple sinks plugged to the decoder (display
and v4l2 for example), the blocking call used for pacing the decoder
output becomes unacceptable, so just remove this useless "feature".
Double-click on extra mouse button to open the settings panel (a
single-click opens the notification panel).
This is consistent with the keyboard shortcut MOD+n+n.
PR #2264 <https://github.com/Genymobile/scrcpy/pull/2264>
Signed-off-by: Romain Vimont <rom@rom1v.com>
The collapsing action collapses any panels.
By the way, the Android method is named collapsePanels().
PR #2260 <https://github.com/Genymobile/scrcpy/pull/2260>
Signed-off-by: Romain Vimont <rom@rom1v.com>
The shortcut "back on screen on" is a bit special: the control is
requested by the client, but the actual event injection (POWER or BACK)
is determined on the device.
To properly inject DOWN and UP events for BACK, transmit the action as
a control parameter.
If the screen is off:
- on DOWN, inject POWER (DOWN and UP) (wake up the device immediately)
- on UP, do nothing
If the screen is on:
- on DOWN, inject BACK DOWN
- on UP, inject BACK UP
A corner case is when the screen turns off between the DOWN and UP
event. In that case, a BACK UP event will be injected, so it's harmless.
As a consequence of this change, the BACK button is now handled by
Android on mouse released. This is consistent with the keyboard shortcut
(Mod+b) behavior.
PR #2259 <https://github.com/Genymobile/scrcpy/pull/2259>
Refs #2258 <https://github.com/Genymobile/scrcpy/pull/2258>
The screen receives callbacks from the decoder, fed by the stream.
The decoder is run from the stream thread, so waiting for the end of
stream is sufficient to avoid possible use-after-destroy.
When --no-display was passed, screen_destroy() was called while
screen_init() was never called.
In practice, it did not crash because it just freed NULL pointers, but
it was still incorrect.
A skipped frame is detected when the producer offers a frame while the
current pending frame has not been consumed.
However, the producer (in practice the decoder) is not interested in the
fact that a frame has been skipped, only the consumer (the renderer) is.
Therefore, notify frame skip via a consumer callback. This allows to
manage the skipped and rendered frames count at the same place, and
remove fps_counter from decoder.
As soon as the stream is started, the video buffer could notify a new
frame available.
In order to pass this event to the screen without race condition, the
screen must be initialized before the screen is started.
Video buffer is a tool between a frame producer and a frame consumer.
For now, it is used between a decoder and a renderer, but in the future
another instance might be used to swscale decoded frames.
It makes sense to extract default values for bitrate and port range
(which are arbitrary and might be changed in the future).
However, the default values for "max size" and "lock video orientation"
are naturally unlimited/unlocked, and will never be changed. Extracting
these options just added complexity for no benefit, so hardcode them.
After the struct screen is initialized, the window, the renderer and the
texture are necessarily valid, so there is no need to check in
screen_destroy().
There were only two frames simultaneously:
- one used by the decoder;
- one used by the renderer.
When the decoder finished decoding a frame, it swapped it with the
rendering frame.
Adding a third frame provides several benefits:
- the decoder do not have to wait for the renderer to release the
mutex;
- it simplifies the video_buffer API;
- it makes the rendering frame valid until the next call to
video_buffer_take_rendering_frame(), which will be useful for
swscaling on window resize.
