Apparently users prefer this behavior.
It was used for subtitles too, so move the code to calculate the video
offset into a separate function. Seeking also needs to be fixed.
Fixes#1018.
The previous commit broke these things, and fixing them is separate in
this commit in order to reduce the volume of changes.
Move the image queue from the VO to the playback core. The image queue
is a remnant of the old way how vdpau was implemented, and increasingly
became more and more an artifact. In the end, it did only one thing:
computing the duration of the current frame. This was done by taking the
PTS difference between the current and the future frame. We keep this,
but by moving it out of the VO, we don't have to special-case format
changes anymore. This simplifies the code a lot.
Since we need the queue to compute the duration only, a queue size
larger than 2 makes no sense, and we can hardcode that.
Also change how the last frame is handled. The last frame is a bit of a
problem, because video timing works by showing one frame after another,
which makes it a special case. Make the VO provide a function to notify
us when the frame is done, instead. The frame duration is used for that.
This is not perfect. For example, changing playback speed during the
last frame doesn't update the end time. Pausing will not stop the clock
that times the last frame. But I don't think this matters for such a
corner case.
The VO is run inside its own thread. It also does most of video timing.
The playloop hands the image data and a realtime timestamp to the VO,
and the VO does the rest.
In particular, this allows the playloop to do other things, instead of
blocking for video redraw. But if anything accesses the VO during video
timing, it will block.
This also fixes vo_sdl.c event handling; but that is only a side-effect,
since reimplementing the broken way would require more effort.
Also drop --softsleep. In theory, this option helps if the kernel's
sleeping mechanism is too inaccurate for video timing. In practice, I
haven't ever encountered a situation where it helps, and it just burns
CPU cycles. On the other hand it's probably actively harmful, because
it prevents the libavcodec decoder threads from doing real work.
Side note:
Originally, I intended that multiple frames can be queued to the VO. But
this is not done, due to problems with OSD and other certain features.
OSD in particular is simply designed in a way that it can be neither
timed nor copied, so you do have to render it into the video frame
before you can draw the next frame. (Subtitles have no such restriction.
sd_lavc was even updated to fix this.) It seems the right solution to
queuing multiple VO frames is rendering on VO-backed framebuffers, like
vo_vdpau.c does. This requires VO driver support, and is out of scope
of this commit.
As consequence, the VO has a queue size of 1. The existing video queue
is just needed to compute frame duration, and will be moved out in the
next commit.
Achieve this by polling. Will be used by the OSC. Basically a bad hack -
but the point is that the mpv core itself is in the best position to
improve this later.
Basically move the code from playloop.c to video.c. The new function
write_video() now contains the code that was part of run_playloop().
There are no functional changes, except handling "new_frame_shown"
slightly differently. This is done so that we don't need new a new
MPContext field or a return value for write_video() to signal this
condition. Instead, it's handled indirectly.
This also reduces some code duplication with other parts of the code.
The changfe is mostly cosmetic, although there are also some subtle
changes in behavior. At least one change is that the big desync message
is now printed after every seek.
Regression since commit 261506e3. Internally speaking, playback was
often not properly terminated, and the main part of handle_keep_open()
was just executed once, instead of any time the user tries to seek. This
means playback_pts was not set, and the "current time" was determined by
the seek target PTS.
So fix this aspect of video EOF handling, and also remove the now
unnecessary eof_reached field.
The pause check before calling pause_player() is a lazy workaround for
a strange event feedback loop that happens on EOF with --keep-open.
If you for example use --audio-file, disable the external track, seek,
and enable the external track again, the playback position of the
external file was off, and you would get major A/V desync. This was
actually supposed to work, but broke at some time ago (probably commit
2b87415f). It didn't work, because it attempted to seek the stream if it
was already selected, which was always true due to
reselect_demux_streams() being called before that.
Fix by putting the initial selection and the seek together.
This commit makes audio decoding non-blocking. If e.g. the network is
too slow the playloop will just go to sleep, instead of blocking until
enough data is available.
For video, this was already done with commit 7083f88c. For audio, it's
unfortunately much more complicated, because the audio decoder was used
in a blocking manner. Large changes are required to get around this.
The whole playback restart mechanism must be turned into a statemachine,
especially since it has close interactions with video restart. Lots of
video code is thus also changed.
(For the record, I don't think switching this code to threads would
make this conceptually easier: the code would still have to deal with
external input while blocked, so these in-between states do get visible
[and thus need to be handled] anyway. On the other hand, it certainly
should be possible to modularize this code a bit better.)
This will probably cause a bunch of regressions.
Broken by commit 1301a907. This commit added demuxer threading, and
changed some other things to make them simpler and more orthogonal. One
of these things was ntofications about streams that appear during
playback. That's an obscure corner case, but the change made handling of
it as natural as normal initialization.
This didn't work for two reasons:
1. When playing an ordered chapters file where the initial segment was
not from the main file, its streams were added to the track list. So
they were printed twice, and switching to the next segment didn't work,
because the right streams were not selected.
2. EDL, CUE, as well as possibly certain Matroska files don't have any
data or tracks in the "main" demuxer, so normally the first segment is
picked for the track list. This was simply broken.
Fix by sprinkling the code with various hacks.
In my opinion this is not really necessary, since there's only a single
user of update_video(), but others reading this code would probably hate
me for using magic integer values instead of symbolic constants.
This should be a purely cosmetic commit; any changes in behavior are
bugs.
Mouse cursor handling, --heartbeat-cmd, and OSD messages basically
relied on polling. For this reason, the playloop always used a small
timeout (not more than 500ms).
Fix these cases, and raise the timeout to 100 seconds. There is no
reason behind this number; for this specific purpose it's as close to
infinity as any other number.
On MS Windows, or if vo_sdl is used, the timeout remains very small.
In these cases the GUI code doesn't do proper event handling in the
first place, and fixing it requires much more effort.
getch2_poll() still does polling, because as far as I'm aware no event-
based way to detect this state change exists.
This adds a thread to the demuxer which reads packets asynchronously.
It will do so until a configurable minimum packet queue size is
reached. (See options.rst additions.)
For now, the thread is disabled by default. There are some corner cases
that have to be fixed, such as fixing cache behavior with webradios.
Note that most interaction with the demuxer is still blocking, so if
e.g. network dies, the player will still freeze. But this change will
make it possible to remove most causes for freezing.
Most of the new code in demux.c actually consists of weird caches to
compensate for thread-safety issues (with the previously single-threaded
design), or to avoid blocking by having to wait on the demuxer thread.
Most of the changes in the player are due to the fact that we must not
access the source stream directly. the demuxer thread already accesses
it, and the stream stuff is not thread-safe.
For timeline stuff (like ordered chapters), we enable the thread for the
current segment only. We also clear its packet queue on seek, so that
the remaining (unconsumed) readahead buffer doesn't waste memory.
Keep in mind that insane subtitles (such as ASS typesetting muxed into
mkv files) will practically disable the readahead, because the total
queue size is considered when checking whether the minimum queue size
was reached.
It's better to keep the logic in one place.
Also drop that a broken config file aborts loading of the player. I
don't see much reason for this, and it inflates the code slightly.
Basically, this allows gapless playback with similar files (including
the ordered chapter case), while still being robust in general.
The implementation is quite simplistic on purpose, in order to avoid
all the weird corner cases that can occur when creating the filter
chain. The consequence is that it might do not-gapless playback in
more cases when needed, but if that bothers you, you still can use
the normal gapless mode.
Just using "--gapless-audio" or "--gapless-audio=yes" selects the old
mode.
mpv_destroy() should perhaps better be called mpv_detach(), because it
destroys only the handle, not necessarily the player. The player is only
terminated if a quit command is sent.
This function quits automatically, and additionally waits until the
player is completely destroyed. It removes the possibility that the
player core is still uninitializing, while all client handles are
already destroyed. (Although in practice, the difference is usually not
important.)
Change how the video decoding loop works. The structure should now be a
bit easier to follow. The interactions on format changes are (probably)
simpler. This also aligns the decoding loop with future planned changes,
such as moving various things to separate threads.
And slightly adjust the semantics of MPV_EVENT_PAUSE/MPV_EVENT_UNPAUSE.
The real pause state can now be queried with the "core-idle" property,
the user pause state with the "pause" property, whether the player is
paused due to cache with "paused-for-cache", and the keep open event can
be guessed with the "eof-reached" property.
This property is set to "yes" if playback was paused due to --keep-open.
The change notification might not always be perfect; maybe that should
be improved.
And consistently use MP_NOPTS_VALUE as error value for the users of this
function. This is better than using -1, especially because negative
values can be valid timestamps.
Remove the ao_buffer_playable_samples field. This contained the number
of samples that fill_audio_out_buffers() wanted to write to the AO (i.e.
this data was supposed to be played at some point), but ao_play()
rejected it due to partial fill.
This could happen with many AOs, notably those which align all written
data to an internal period size (often called "outburst" in the AO
code), and the accepted number of samples is rounded down to period
boundaries. The left-over samples at the end were still kept in
mpctx->ao_buffer, and had to be played later.
The reason ao_buffer_playable_samples had to exist was to make sure that
at EOF, the correct number of left-over samples was played (and not
possibly other data in the buffer that had to be sliced off due to
endpts in fill_audio_out_buffers()). (You'd think you could just slice
the entire buffer, but I suspect this wasn't done because the end time
could actually change due to A/V sync changes. Maybe that was the reason
it's so complicated.)
Some commits ago, ao.c gained internal buffering, and ao_play() will
never return partial writes - as long as you don't try to write more
samples than ao_get_space() reports. This is always the case. The only
exception is filling the audio buffers while paused. In this case, we
decode and play only 1 sample in order to initialize decoding (e.g. on
seeking). Actually playing this 1 sample is in fact a bug, but even of
the AO doesn't have period size alignment, you won't notice it. In
summary, this means we can safely remove the code.
We want to move the AO to its own thread. There's no technical reason
for making the ao struct opaque to do this. But it helps us sleep at
night, because we can control access to shared state better.
For example, consider the case when audio initialization fails. Then the
audio track is deselected. Before this commit, this would have been
equivalent to the user disabling audio. This is bad when multiple files
are played at once (the next file would have audio disabled, even if it
works), or if playback resume is used (if e.g. audio output failed to
initialize, then audio would be disabled when resuming, even if the
system's audio driver was fixed).
The OSC used significant CPU time while the player was paused. It turned
out that the "tick" event sent during pause is the problem. The OSC
accesses the player core when receiving a tick event, which in turn will
cause the core to send another tick event, leading to infinite feedback.
Fix this by sending an idle tick only every 500ms. This is not very
proper, but the idea behind the tick event isn't very clean to begin
with (and the OSC should use timers instead).
This is approximate: we read each option value on program start
(before starting playback of a file), and when writing the resume
config, compare each value to the current state. This also means
when a value is changed and then changed back, it's not stored. In
particular, option values set in config files and on the command
line are considered the default.
This should help reducing the numbers of options overridden by the
resume config. If too much is overridden, it becomes an inconvenience,
because changes in config files will apparently have no effect when
resuming a file.
Also see github issue #574.
Not sure about this... might redo.
At least this provides a case of a broadcasted event, which requires
per-event data allocation.
See github issue #576.
The initialization code was split and refactored for the libmpv changes.
One change, moving a part of cocoa initialization, accidentally broke
--force-window on OSX, which creates a VO in a certain initialization
stage. We still don't know how cocoa should behave with libmpv, so fix
this with a hack to beat it back into working. Untested.
This library will export the client API functions.
Note that this doesn't allow compiling the command line player to link
against this library yet. The reason is that there's lots of weird stuff
required to setup the execution environment (mostly Windows and OSX
specifics), as well as things which are out of scope of the client API
and every application has to do on its own. However, since the mpv
command line player basically reuses functions from the mpv core to
implement these things, it's not very easy to separate the command
line player form the mpv core.
This is partial only, and it still accesses some MPContext internals.
Specifically, chapter and track lists are still read directly, and OSD
access is special-cased too.
The OSC seems to work fine, except using the fast-forward/backward
buttons. These buttons behave differently, because the OSC code had
certain assumptions how often its update code is called.
The Lua interface changes slightly.
Note that this has the odd property that Lua script and video start
at the same time, asynchronously. If this becomes an issue, explicit
synchronization could be added.
Add a client API, which is intended to be a stable API to get some rough
control over the player. Basically, it reflects what can be done with
input.conf commands or the old slavemode. It will replace the old
slavemode (and enable the implementation of a new slave protocol).
The code removed from handle_input_and_seek_coalesce() did two things:
1. If there's a queued seek, stop accepting non-seek commands, and delay
them to the next playloop iteration.
2. If a seek is executing (i.e. the seek was unqueued, and now it's
trying to decode and display the first video frame), stop accepting
seek commands (and in fact all commands that were queued after the
first seek command). This logic is disabled if seeking started longer
than 300ms ago. (To avoid starvation.)
I'm not sure why 1. would be needed. It's still possible that a command
immediately executed after a seek command sees a "seeking in progress"
state, because it affects queued seeks only, and not seeks in progress.
Drop this code, since it can easily lead to input starvation, and I'm
not aware of any disadvantages.
The logic in 2. is good to make seeking behave much better, as it
guarantees that the video display is updated frequently. Keep the core
idea, but implement it differently. Now this logic is applied to seeks
only. Commands after the seek can execute freely, and like with 1., I
don't see a reason why they couldn't. However, in some cases, seeks are
supposed to be executed instantly, so queue_seek() needs an additional
parameter to signal the need for immediate update.
One nice thing is that commands like sub_seek automatically profit from
the seek delay logic. On the other hand, hitting chapter seek multiple
times still does not update the video on chapter boundaries (as it
should be).
Note that the main goal of this commit is actually simplification of the
input processing logic and to allow all commands to be executed
immediately.
Do two things:
1. add locking to struct osd_state
2. make struct osd_state opaque
While 1. is somewhat simple, 2. is quite horrible. Lots of code accesses
lots of osd_state (and osd_object) members. To make sure everything is
accessed synchronously, I prefer making osd_state opaque, even if it
means adding pretty dumb accessors.
All of this is meant to allow running VO in their own threads.
Eventually, VOs will request OSD on their own, which means osd_state
will be accessed from foreign threads.
These were needed before the last commit, but now they don't do anything
anymore. (They were used to decide whether to replace or stack the
previous OSD message when a new one was displayed.)
If certain OSD messages were displayed at the same time, the hidden
messages were put on the stack, and displayed again once the higher
priority messages disappeared. The idea was probably that lower priority
messages could not hide higher priority ones, and also that the lower
messages did not get lost.
But in practice, this gives confusing results with OSD messages randomly
reappearing for a brief time. Remove it.
Showing subtitles on terminal used the OSD message stack (which uses a
stack to "pile up" messages that were displayed at the same time). This
had a bunch of weird and annoying consequences. This accessed a certain
osd_state field, which is a minor annoyance since I want to make that
struct opaque. Implement this differently.
wm4
Andrey Morozov
Tsukasa OMOTO
xylosper