If the audio callback suddenly stops, and the AO provides no "reset"
callback, then reset() could deadlock by waiting on the audio callback
forever.
The waiting was needed to enter a consistent state, where the audio
callback guarantees it won't access the ringbuffer. This in turn is
needed because mp_ring_reset() is not concurrency-safe.
This active waiting is unavoidable. But the way it was implemented, the
audio callback had to call ao_read_data() at least once when reset() is
called. Fix this by making ao_read_data() set a flag upon entering and
leaving, which basically turns p->state into some sort of spinlock.
The audio callback actually never needs to spin, because there are only
2 states: playing audio, or playing silence. This might be a bit
surprising, because usually atomic_compare_exchange_strong() requires a
retry-loop idiom for correct operation.
This commit is needed because ao_wasapi can (or will in the future)
randomly stop the audio callback in certain corner cases. Then the
player would hang forever in reset().
As usual, we use C11 semantics, and emulate it if <stdatomic.h> is not
available.
It's a bit messy with __sync_val_compare_and_swap(). We assume it has
"strong" semantics (it can't fail sporadically), but I'm not sure if
this is really the case. On the other hand, weak semantics don't seem to
be possible, since the builtin can't distinguish between the two failure
cases that could occur. Also, to match the C11 interface, use of gcc
builtins is unavoidable. Add a check to the build system to make sure
the compiler supports them (although I don't think there's any compiler
which supports __sync_*, but not these extensions).
Needed for the following commit.
This is what you would expect. Before this commit, each
ao_request_reload() call would just queue a reload command, and then
recreate the AO for the number of times the function was called.
Instead of sending a command, introduce some sort of event retrieval
mechanism. At least for the reload case, use atomics, because we're too
lazy to setup an extra mutex.
This commit fixes a "cosmetic" user interface issue. Instead of
displaying the interpolated seek time on OSD, show the actual audio
time.
This is rather silly: when seeking in audio-only mode, it takes some
iterations until audio is "ready", but on the other hand, the audio
state machine is rather fickle, and fixing this cosmetic issue would be
intrusive. So just add a hack that paints over the ugly behavior as
perceived by the user. Probably the lesser evil.
It doesn't happen if video is enabled, because that mode sets the
current time immediately to video PTS. (Audio has to be synced to video,
so the code is a bit more complex.)
Fixes#1233.
This reverts commit d859549424.
Going to apply the alternative fix through PR #1256, which came just
some seconds after pushing the reverted commit. The reverted commit
was reported as not actually working.
Found by clang sanitizer. Casting unsigned integers to signed integers
with same size has implementation defined behavior (it's even allowed to
crash), but it seems reasonable to expect that reasonable
implementations do a complement of 2 "conversion".
Following the discussion in #1253.
The events won't be removed for a while, though. (Or maybe never, unless
we run out of bits for the uint64_t event mask.)
This is not a real change (the events still work, and the alternative
mechanisms were established a few API revisions earlier), but for the
sake of notifying API users, update DOCS/client-api-changes.rst.
The values compared here happen to be of unsigned enum types - but the
test is not supposed to break if we somehow force the enum to signed, or
if the compiler happens to use a signed type (as far as I remember, the
exact integer type the compiler can use is implementation-defined).
The main need I see for this is with libmpv - it would be confusing if
some application showed up as "mpv" on whateverthehell PulseAudio uses
it for (generally it does show up on various PA GUI tools).
This silences the warning:
video/out/gl_video.c:1091:51: runtime error: division by zero
when running with clang -fsanitize=undefined. Division by zero is legal
according to IEEE, but I guess clang doesn't care about standard. While
triggering this warning isn't actually avoided in all cases, it's
avoided in the common case and also makes people shut up about it.
Call VOCTRL_GET_DISPLAY_NAMES it when the property is
requested. The vo should return the names of the displays that the mpv
window is covering. For example, with x11 vos, xrandr names LVDS1,
HDMI1, etc.
XRRGetOutputInfo contains a "name" element which corresponds to to the
display names given to the user by the "xrandr" command line
utility. Copy it into the xrandr_display struct for each display.
On VOCTRL_GET_DISPLAY_NAMES, send a copy of the names
of the displays spanned by the mpv window on.
The intention is to avoid using the timeout-based fallback.
There's some minor hope that this will help with OpenBSD (see #1239),
although it probably won't.
Some chance that this will cause trouble with obscure OSS
implementations or emulations.
If calling ao->driver->wait() fails, we need to fallback to timeout-
based waiting. But it could be that at this point, the mutex was already
released (and then re-acquired). So we need to recheck the condition in
order to avoid missed wakeups.
This probably wasn't an actually occurring problem, but still could
cause a small race-condition window if the dynamic fallback is actually
used.
This considered only index entries that were for the same track ID as
the track used for seeking. This doesn't make much sense for preroll;
it'll just possibly skip clusters, and select an earlier cluster.
One possible negative side-effect is that the preroll might be too tight
now, and miss subtitle packets more often.
The demuxer has a hack to seek to the cluster before the target cluster
in order to "catch" subtitle lines that start before the seek target,
but overlap with the video after the seek target.
Avoid this hack if the cue index indicates that there are no overlapping
subtitle packets that can be caught by seeking to the previous cluster.
Nothing is done with them yet. This is preparation for the following
commit.
CueRelativePosition isn't even saved anywhere, because I don't intend to
use it. (Too messy for no gain.)
Instead of indexing only 1 packet per cluster (which is enough for
working seeking), add every packet to the index.
Since on seek, we go through every single index entry, this probably
makes seeking slower. On the other hand, this code is used for files
without index only (e.g. incomplete files), so it probably doesn't
matter much.
Preparation for the following commits.
update_subtitle() already uees playback_pts to make subtitles work
better in no-audio mode. Using get_current_time() usually gets
playback_pts, but also has the advantage that it will use the seek
target time during seeks. This will result in multiple sub_seek commands
doing the right thing (at least as long as they're far enough apart so
that seeking is actually initiated when the second command is run).
If no packets are queued, the readahead time is obviously 0.
If the end time is smaller than the start time, the problem is probably
that audio and video start at slightly different times - report 0 in
this case too.
Do this because seeing "???" as readahead time is a bit annoying.
Instead, use the native-endian alias, and switch the wayland format
depending on the target platform's endian.
This drops support for swapped-endian formats, but I think that is ok.
Not only are the affected formats rather ancient and backwards, but
using swapped formats probably does not make any sense for performance
either.
Untested.
These formats are still supported; you just can't reference them via a
defined constants directly. They are now handled via the generic
passthrough.
(If you want to use such a format, you either have to add the entry
back, or use AV_PIX_FMT_* directly.)
This is a rather radical change: instead of maintaining a whitelist of
FFmpeg formats we support, we automatically support all formats.
In general, a format which doesn't have an explicit IMGFMT_* name will
be converted to a known format through libswscale, or will be handled
by code which can treat pixel formats in a generic way using the pixel
format description, like vo_opengl.
AV_PIX_FMT_UYYVYY411 is a special-case. It's packed YUV with chroma
subsampling by 4 in both directions. Its component order is documented
as "Cb Y0 Y1 Cr Y2 Y3", meaning there's one UV sample for 4 Y samples.
This means each pixel uses 1.5 bytes (4 pixels have 1 UV sample, so
4 bytes + 2 bytes). FFmpeg can actually handle this format with its
generic mechanism in an extremely awkward way, but it doesn't work for
us. Blacklist it, and hope no similar formats will be added in the
future.
Currently, the AV_PIX_FMT_*s allowed are limited to a numeric value of
500. More is not allowed, and there are some fixed size arrays that need
to contain any possible format (look for IMGFMT_END dependencies).
We could have this simpler by replacing IMGFMT_* with AV_PIX_FMT_*
through the whole codebase. But for now, this is better, because we
can compensate for formats missing in Libav or older FFmpeg versions,
like AV_PIX_FMT_RGB0 and others.
FFmpeg has only a AV_PIX_FMT_FLAG_BE flag, not a LE one, which causes
problems for us: we want to have the LE flag too, so code can actually
detect whether a format is non-native endian. Basically, we want to
reconstruct the LE/BE suffix all AV_PIX_FMT_*s have.
Doing this is hard due to the (messed up) way AVPixFmtDescriptor works.
The worst is AV_PIX_FMT_RGB444: this group of formats describe an
endian-independent access (since no component actually spans 2 bytes,
you only need byte accesses with a fixed offset), so we have to go
through some pain.
This influences the demuxer readahead display. If a stream has reached
EOF, we want to ignore it for the purpose of this calculation.
Note that if a stream contains no packets, it still should cause the
value 0s to be displayed (unless it's EOF), because that's just the
actual situation.
