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.
The function video_decode_and_filter(), called between initializing the
local vf variable and using it, can actually destroy and recreate the
filter. Thus, the vf variable turns into a dangling pointer if that
happens.
Could be observed with: --hwdec=vda --deinterlace=yes --vf=yadif
(Also happens with vdpau/vaapi.)
This code was sending a string to a different thread, and then
deallocated the string shortly after, which means most of the time
the other thread was accessing a dangling pointer.
It's possible that this is the cause for #1002.
Completely useless, and could accidentally be enabled by cycling
framedrop modes. Just get rid of it.
But still allow triggering the old code with --vd-lavc-framedrop, in
case someone asks for it. If nobody does, this new option will be
removed eventually.
Trying to jump chapters in a gile that has no chapters does nothing,
not even show a warning. This is confusing. The reason is that the
"add chapter" command will just bail out completely if the property
is unavailable.
This was because it exited when it couldn't get the property type.
Instead of exiting, just don't enter the code that needs the type.
(I'm not sure when this behavior changed. I consider it a regression.
It was probably caused by changes to the chapter code, which perhaps
started returning UNAVAILABLE instead of OK if there are no chapters.)
The client API exports this state via events already, but maybe it's
better to explicitly provide this property in order to facilitate use on
OSD and similar cases.
This is delayed by 300ms - before that, the status doesn't change. I
feel like it would too annoying if the status line would "flicker" on
normal seek by quickly showing and hiding the indicator.
If this code is not skipped, encoding (or dumping with --ao=pcm) will
attempt to adjust video timing to audio. Since another commit (0cce8fe6)
already avoids writing audio ahead, this didn't slow down encoding to
realtime, but it was still significantly slower.
This change should actually remove all extra sleeping.
This builds but both the libmpv example and the cplayer block infinitely when
building libmpv. That's because we wait inifinitely in `dispatch_sync` as
there's no event loop in the main thread that allows for libdispatch to work..
Whiel we are at it, we should probably investigate how to use mp_dispatch
instead since it is a little lower level and could give us higher control in
building and event loop.
The subtitle timing logic always used the demuxer's start time as video
offset. This made external subtitle files "just work" with file formats
like TS, which usually have a non-0 start time. But it was wrong for
subtitles muxed with the TS, so adjust the time offset explicitly with
external files only.
Although disabling both video and audio is surely an obscure corner
case, it's allowed, and we don't want the demuxer to skip arbitrary
packets.
Basically, make the heuristic for checking interleaved files affect
external files only.
Handle --term-playing-msg at a better place.
Move MPV_EVENT_TICK hack into a separate function. Also add some words
to the client API that you shouldn't use it. (But better leave breaking
it for later.)
Handle --frames and frame_step differently. Remove the mess from the
playloop, and do it after frame display. Give up on the weird semantics
for audio-only mode (they didn't make sense anyway), and adjust the
manpage accordingly.
If seeks take very long, it's better not to freeze up the display.
(This doesn't handle the case when decoding video frames is extremely
slow; just if hr-seek is used, or the demuxer is threaded and blocks on
network I/O.)
Internally, there are two mechanisms which can trigger property
notification as used with "observed" properties in the client API.
The first mechanism associates events with a group of properties that
are potentially changed by a certain event. mp_event_property_change[]
declares these associations, and maps each event to a set of strings.
When an event happens, the set of strings is matched against the list of
observed properties of each client. Make this more efficient by
comparing bitsets of events instead. This way, only a bit-wise "and" is
needed for each observed property. Even better, we can completely skip
clients which have no observed properties that match.
The second mechanism just updates individual properties explicitly by
name. Optimize this by using the property index instead. It would be
nice if we could reuse the first mechanism for the second one, but
there are too many properties to fit into a 64 bit mask.
(Though the limit on 64 events might get us into trouble later...)
Almost nothing was left of it.
The only thing this commit actually removes is support for reading
input commands from stdin. But you can emulate this via:
--input-file=/dev/stdin --input-terminal=no
However, this won't work on Windows. Just use a named pipe.
Playing audio files with embedded cover art broke due to some of the
recent changes. Treat video EOF properly, and don't burn the CPU.
Disable hrseek for video in attached picture mode, since the decoder
will always produce a new image, which makes hrseek never terminate.
Fixes#970.
In encoding mode, the AO pretends to be infinitely fast (it will take
whatever we write, without ever rejecting input). Commit 261506e3 broke
this somehow. It turns out an old hack dealing with this was accidentally
dropped.
This is the hunk of code whose semantics were (partially) dropped:
if (mpctx->d_audio && (mpctx->restart_playback ? !video_left :
ao_untimed(mpctx->ao) && (mpctx->delay <= 0 ||
!video_left)))
{
int status = fill_audio_out_buffers(mpctx, endpts);
// Not at audio stream EOF yet
audio_left = status > -2;
}
This if condition is pretty wild, and it looked like it was pretty much
for audio-only mode, rather than subtle handling for encoding mode.
"Internal" events were added in the previous commits to leverage the
client API property mechanism, without making weird properties public.
But they were sent to clients too (and returned by mpv_wait_event()).
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.
This is perfectly allowed, but was ignored, because it's a corner case.
It doesn't actually wait for other clients to be destroyed, but on the
other hand I think there's no way to have other clients before
initialization.
CC: @mpv-player/stable
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.
Frames buffered in filters weren't flushed, so on EOF, the last frames
were dropped, depending on how much filters buffered. Oops.
Test case: "mpv something.jpg --vf=buffer"
In situations when the demuxer reports EOF, but immediately "recovers"
after that and returns new data, it could happen that audio sync was
skipped. Deal with this by actually entering the EOF state, instead of
assuming this will happen later.
Some files have the first audio much later into the video (for whatever
reasons). Instead of appending large amounts of silence to the audio
buffer (and refusing to sync if the audio to append is "too large"),
just wait until enough video has played.
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 an imprecise seek is issues while a precise seek is ongoing,
don't wait up to 300ms (herustistic which usually improves user
experience), but instead let it cancel the seek.
Improves responsiveness of the OSC after the previous commit.
Note that we don't do this on "default-precise" seeks, because we
don't know if they're going to be precise or not.
It probably happens relatively often that the first packet (or even the
first N packets) of a stream will fail to decode, but decoding will
eventually succeed at a later point. Before commit 261506e3, this was
handled by an explicit retry loop (although this was also for other
purposes), but with then was changed to abort on the first error. This
makes it impossible to decode some audio streams.
Change this so that errors are ignored for the first 50 packets, which
should make it equivalent to the old code.
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.
Seeking in .ts files (and some other formats) is too unreliable, so
there's a separate code path for this case. But it breaks hr-seek.
Maybe hr-seek could actually be enabled in this case if we're careful
enough about timestamp resets, but for now nothing changes.
