When initialization failed, vo_lavc may cause an irrecoverable state in
the ffmpeg-related structs. Therefore, we reject additional
initialization attempts at least until we know a better way to clean up
the mess.
ao_lavc currently cannot be initialized more than once, yet it's good to
do consistent changes there as well.
Also, clean up uninit-after-failure handling to be less spammy.
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().
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.
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).
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.
Apparently this can "sometimes" return an error. In my opinion, this
should never return an error: neither the semantics of the function,
nor the ALSA documentation or ALSA sample code seem to indicate that
a failure is to be expected. I'm not perfectly sure about this though
(I blame ALSA being a weird, big, underdocumented API).
Since it causes problems for some users, and since there is really no
reason why we should abort on such an error, turn it into a warning.
Fixes#1231.
Since the list associated with --audio-device is supposed to enable
simple user-selection, it doesn't make much sense to include overly
special things like ao_pcm or ao_null in the list. Specifically,
ao_pcm is harmful, because it will just dump all audio to a file
named audiodump.wav in the current working directory. The user can't
choose the filename (it can be customized, but not through this
option), and the working directory might be essentially random,
especially if this is used from a GUI.
Exclude "strange" entries. We reuse the fact that there's already a
simple list ordered by auto-probe priority in order to avoid having to
add an additional flag. This is also why coreaudio_exclusive was moved
above ao_null: ao_null ends auto-probing and marks the start of
"special" outputs, which don't show up on the device, but we want
coreaudio_exclusive to be selectable (I think).
Move it above ao_null, so that it can be selected during auto-probing
(even if it's only last). I see no reason why it should not be included,
and it makes the following commit slightly more elegant. (See
explanations there.)
Especially with other components (libavcodec, OSX stuff), the thread
list can get quite populated. Setting the thread name helps when
debugging.
Since this is not portable, we check the OS variants in waf configure.
old-configure just gets a special-case for glibc, since doing a full
check here would probably be a waste of effort.
While conceptually this sink stuff in PulseAudio does just the right
thing, actually listing the sinks is unbelievable complicated. Not only
is the idea that listing them should happen asynchronously completely
bullshit (who the fuck runs the PulseAudio server on a separate
computer), but the way this is done is full of bullshit too. Why
separate callbacks for each device? Why this obtuse mainloop shit?
Especially the mainloop shit makes it actively worse than doing things
manually with pthread primitives, and the reason for that (different
mainloop implementations for GUIs?) is laughable too. It's like they
chose the most complicated API possible just because they attempted
to "abstract" basic mechanisms in order to handle "everything". While
I don't claim to design the best APIs, this API is fucking terrible
without any excuse. (End of rant.)
All the dumb crap in pa_init_boilerplate() is needed to talk to the
audio server at all. Might also fix some subtle bugs in the init code
(which is strange, because the original file was contributed by the
devil himself).
The one in msg.c was mistakenly removed with commit e99a37f6.
I didn't actually test the change in ao_sndio.c (but obviously "ap"
shouldn't be static).
Don't wait after the audio thread has pushed the remaining audio to the
AO. Avoids hard hangs if the heuristic fails completely (could still
happen if get_delay returns absurd values).
CC: @mpv-player/stable
Since the internal AO driver API has no proper way to determine EOF, we
need to guess by querying get_delay. But some AOs (e.g. ao_pulse with
no-latency-hacks set) may never reach 0, maybe because they naively add
the latency to the buffer level. In this case our heuristic can break.
Fix by always using the delay to estimate the EOF time. It's not even
that important - it's mostly used to avoid blocking draining. So this
should be ok.
CC: @mpv-player/stable (maybe)
Unfortunately, ALSA is particularly bad with this, because mpv has to
add all sorts of magic crap to the device name to make things work. The
device selection overrides this, so explicitly selecting devices will
most likely break your audio. This has yet to be solved.
This function is available starting with PulseAudio 2.0, while we only
require 1.0. This broke compilation on Ubuntu 12.04.5 LTS.
Use our own function to calculate the buffer size, which is actually
simpler and needs slightly less code.
Hopefully fixes#1154.
CC: @mpv-player/stable
It was more complicated than it had to be: the audio thread already
determines whether audio has ended, so we can use that. Remove the
separate logic for draining.
Commit 957097 attempted to use PA_STREAM_FAIL_ON_SUSPEND to make
ao_pulse exit if the stream was started suspended.
Unfortunately, PA_STREAM_FAIL_ON_SUSPEND is active even during playback.
If you pause mpv, pulseaudio will close the actual audio device after a
while (or something like this), and unpausing won't work. Instead, it
will spam "Entity killed" error messages.
Undo this change and check for suspended audio manually during init.
CC: @mpv-player/stable
Sometimes, ao_pulse starts in suspended mode, which means playback is
essentially paused in pulseaudio. This gives the impression that mpv is
hanging, since it times video against the audio playback progress, and
audio never makes progress in this state.
I'm not sure if this will help - possibly it does with mixed
pulseaudio/alsa setups. However, if the alsa setup has the pulseaudio
plugin, alsa will hang too. But there's still a chance we get less
blame for pulseaudio messes.
libsndio has absolutely no mechanism to discard already written audio
(other than SIGKILLing the sound server). sio_stop() will always block
until all audio is played. This is a legitimate design bug.
In theory, we could just not stop it at all, so if the player is e.g.
paused, the remaining audio would be played. When resuming, we would
have to do something to ensure get_delay() returns the right value. But
I couldn't get it to work in all cases.
get_delay needs to report the current audio buffer status. It's
important for A/V sync that this information is current, but functions
which update it were called on play() or get_space() calls only.
This was in bytes, but it's more convenient to use samples (or frames;
in any case the smallest unit of audio that includes all channels).
Remove the ao->bps line too; it will be set after init() returns.
Digital pass-through was probably broken. Possibly fix it (no way to
test). This also should make the logic slightly saner.
Fortunately, it's unlikely that anyone who uses OSS has a spdif setup.
Commit 5b5a3d0c broke this. The really funny thing is that this code was
actually always under "#if BYTE_ORDER == BIG_ENDIAN". The breaking
commit just edited this code slightly, but it must have failed to
compile on big endian long before (since over 1 year ago, commit d3fb58).
Should be able to pass-through AC3, DTS, and others.
It seems PulseAudio wants players to fallback to PCM on certain events
signaled by the server, but we don't implement that. There's not much
documentation available anyway.
Before this commit, there was AF_FORMAT_AC3 (the original spdif format,
used for AC3 and DTS core), and AF_FORMAT_IEC61937 (used for AC3, DTS
and DTS-HD), which was handled as some sort of superset for
AF_FORMAT_AC3. There also was AF_FORMAT_MPEG2, which used
IEC61937-framing, but still was handled as something "separate".
Technically, all of them are pretty similar, but may use different
bitrates. Since digital passthrough pretends to be PCM (just with
special headers that wrap digital packets), this is easily detectable by
the higher samplerate or higher number of channels, so I don't know why
you'd need a separate "class" of sample formats (AF_FORMAT_AC3 vs.
AF_FORMAT_IEC61937) to distinguish them. Actually, this whole thing is
just a mess.
Simplify this by handling all these formats the same way.
AF_FORMAT_IS_IEC61937() now returns 1 for all spdif formats (even MP3).
All AOs just accept all spdif formats now - whether that works or not is
not really clear (seems inconsistent due to earlier attempts to make
DTS-HD work). But on the other hand, enabling spdif requires manual user
interaction, so it doesn't matter much if initialization fails in
slightly less graceful ways if it can't work at all.
At a later point, we will support passthrough with ao_pulse. It seems
the PulseAudio API wants to know the codec type (or maybe not - feeding
it DTS while telling it it's AC3 works), add separate formats for each
codecs. While this reminds of the earlier chaos, it's stricter, and most
code just uses AF_FORMAT_IS_IEC61937().
Also, modify AF_FORMAT_TYPE_MASK (renamed from AF_FORMAT_POINT_MASK) to
include special formats, so that it always describes the fundamental
sample format type. This also ensures valid AF formats are never 0 (this
was probably broken in one of the earlier commits from today).
This code tried to play with the format bits, and potentially could
create invalid formats, or reinterpret obscure formats in unexpected
ways.
Also there was an abort() call if the winapi or mpv used a format with
unexpected bit-width. This could probably easily happen; for example,
mpv supports at least one 64 bit format. And what would happen on 8 bit
formats anyway?
Untested.
Until now, the audio chain could handle both little endian and big
endian formats. This actually doesn't make much sense, since the audio
API and the HW will most likely prefer native formats. Or at the very
least, it should be trivial for audio drivers to do the byte swapping
themselves.
From now on, the audio chain contains native-endian formats only. All
AOs and some filters are adjusted. af_convertsignendian.c is now wrongly
named, but the filter name is adjusted. In some cases, the audio
infrastructure was reused on the demuxer side, but that is relatively
easy to rectify.
This is a quite intrusive and radical change. It's possible that it will
break some things (especially if they're obscure or not Linux), so watch
out for regressions. It's probably still better to do it the bulldozer
way, since slow transition and researching foreign platforms would take
a lot of time and effort.
IEC 61937 frames should always be little endian (little endian 16 bit
words). I don't see any apparent need why the audio chain should handle
swapped-endian formats.
It could be that some audio outputs might want them (especially on big
endian architectures). On the other hand, it's not clear how that works
on these architectures, and it's not even known whether the current code
works on big endian at all. If something should break, and it should
turn out that swapped-endian spdif is needed on any platform/AO,
swapping still could be done in-place within the affected AO, and
there's no need for the additional complexity in the rest of the player.
Note that af_lavcac3enc outputs big endian spdif frames for unknown
reasons. Normally, the resulting data is just pulled through an auto-
inserted conversion filter and turned into little endian. Maybe this was
done as a trick so that the code didn't have to byte-swap the actual
audio frame. In any case, just make it output little endian frames.
All of this is untested, because I have no receiver hardware.
Pausing/unpausing while the audio device can't be reopened, and then
unpausing again when the device is finally reopened, can hang the
player for a while.
This happens because p->prepause_samples grows without bounds each
time the player is unpaused while the device is lost. On unpause,
ao_oss plays prepause_samples of silence to compensate for A/V timing
issues due to the partially lost buffer (we can't pause the device at
an arbitrary sample position, and the current period will be lost).
This in turn will make the player appear to be frozen if too much
audio is queued. (Normally, play() must never block, but here it
happens because more data is written than get_space() reports. A
better implementation would never let prepause_samples grow larger
than the period size.)
The unbounded growth happens because get_space() always returns that
the device can be written while the device is lost. So limit it to
200ms. (A better implementation would limit it to the period size.)
Also see #1080.
Apparently NetBSD users want/need this (see issue #1080).
In order not to break playback, we need at least to emulate get_delay().
We do this approximately by using the system clock.
Also, always close the audio device on reset. Reopen it on play only. If
we can't reopen it, don't retry until after the next time reset or
resume is called, to avoid spam and unexpectedly "stealing" back the
audio device.
Also do something about framestepping causing audio desync.
The context struct had an audio_buf_info field, but there's no reason
why this would be needed. It's a tiny struct, and it isn't permanent
state. It's always returned by SNDCTL_DSP_GETOSPACE. Keeping this as
field is just confusing, so get rid of it.
The code for reopening the audio device was separate, and duplicated
some of the "real" open code. This was very badly done, and major
required parts of initialization were skipped. Fix this by removing
the code duplication. This consists mainly of moving the code for
opening the device to a separate function, and adding some changes
to handle format changes gracefully. (We can't change the audio
format on the fly, but we can at least not explode and play noise
when that happens.)
As a minor change, actually always use SNDCTL_DSP_RESET when closing
the audio device. We don't want to wait until the rest of the buffer
is played.
Also, don't use strerror() when printing the error message that
reopening failed, simply because reopen_device() takes care of this,
and also errno might be clobbered at this point.
I have no idea whether this is true, because there literally doesn't
seem to exist documentation for SNDCTL_DSP_RESET. But at least on
Linux' OSS emulation, it is true. Also, it would be quite insane if
it would be needed.
It seems on NetBSD SNDCTL_DSP_RESET exists, but using it for pausing
is not feasible. We still use it to discard the audio buffer when
closing the audio device.
Replace select() usage with poll() (and reduce code duplication).
Also, while we're at it, drop --disable-audio-select, since it has the
wrong name anyway. And I have doubts that this is needed anywhere. If
it is, it should probably fallback to doing the right thing by default,
instead of requiring the user to do it manually. Since nobody has done
that yet, and since this configure option has been part of MPlayer ever
since ao_oss was added, it's probably safe to say it's not needed.
The '#ifdef SNDCTL_DSP_GETOSPACE' was pointless, since it's already used
unconditionally in another place.
Improve the logic how the audio thread decides how to wait until the AO
is ready for new data. The previous commit makes some of this easier,
although it turned out that it wasn't required, and we still can handle
AOs with bad get_space implementation (although the new code prints an
error message, and it might fail in obscure situations).
The new code is pretty similar to the old one, and the main thing that
changes is that complicated conditions are tweaked. AO waiting is now
used better (mainly instead of max>0, r>0 is used). Whether to wakeup
is reevaluated every time, instead of somehow doing the wrong thing
and compensating for it with a flag.
This fixes the specific situation when the device buffer is full, and
we don't want to buffer more data. In the old code, this wasn't handled
correctly: the AO went to sleep forever, because it prevented proper
wakeup by the AO driver, and as consequence never asked the core for new
data. Commit 4fa3ffeb was a hack-fix against this, and now that we have
a proper solution, this hack is removed as well.
Also make the refill threshold consistent and always use 1/4 of the
buffer. (The threshold is used for situations when an AO doesn't
support proper waiting or chunked processing.)
This commit will probably cause a bunch of regressions again.
Round get_space() results in the same way play() rounds the input size.
Some audio APIs do this for various reasons.
This affects only "push" based AOs. Some of these need no change,
because they either do it already right (like ao_openal), or they seem
not to have any such requirements (like ao_pulse).
Needed for the following commit.
Remove the unnecessary indirection through ao fields.
Also fix the inverted result of AOCONTROL_HAS_TEMP_VOLUME. Hopefully the
change is equivalent. But actually, it looks like the old code did it
wrong.
With --gapless-audio=no, changing from one file to the next apparently
made it hang, until the player was woken up by unrelated events like
input. The reason was that the AO doesn't notify the player of EOF
properly. the played was querying ao_eof_reached(), and then just went
to sleep, without anything waking it up.
Make it event-based: the AO wakes up the playloop if the EOF state
changes.
We could have fixed this in a simpler way by synchronously draining the
AO in these cases. But I think proper event handling is preferable.
Fixes: #1069
CC: @mpv-player/stable (perhaps)
The audio/video sync code in player/audio.c calls ao_reset() each time
audio decoding is entered, but the player is paused, and there would be
more than 1 sample to skip to make audio start match with video start.
This caused a wakeup feedback loop with push.c.
CC: @mpv-player/stable
bstr.c doesn't really deserve its own directory, and compat had just
a few files, most of which may as well be in osdep. There isn't really
any justification for these extra directories, so get rid of them.
The compat/libav.h was empty - just delete it. We changed our approach
to API compatibility, and will likely not need it anymore.
The original intention was probably to avoid unnecessarily high numbers
of wakeups. Change it to wait at most 25% of buffer time instead of 75%
until refilling. Might help with the dsound problems in issue #1024, but
I don't know if success is guaranteed.
Reduce from 1000ms to 100ms. Since there is an audio thread updating AOs
quickly enough now, requesting such a large buffer size makes no sense
anymore.
Some ALSA plugins take non-interleaved audio, but treat it as
interleaved, which results in various funny bugs. Users keep hitting
this issue, and it just doesn't seem worth the trouble.
CC: @mpv-player/stable
Add an option that enables using native PulseAudio auto-updated timing
information, instead of the manual calculations added in mplayer2 times.
You can use --ao=pulse:no-latency-hacks to enable the new code. The code
is almost the same as the code that was removed with commit de435ed5,
but I didn't readd some bits I didn't understand. Likewise, the option
will disable the code added with that commit.
In my tests this seemed to work well, though the A/V sync display looks
funny when seeking.
The default is still the old behavior.
See issue #959.
This was needed by very old (0.9) versions only. Get rid of it.
Unfortunately, I can't cross-check with the original bug report, since
the bug URL leads to this:
Internal Server Error
TracError: IOError: [Errno 2] No such file or directory: '/home/lennart/svn/trac/pulseaudio/VERSION'
ao_null is used to stop autoprobing (if all AOs before fail to init).
After it come things like ao_pcm, which should never be automatically
selected.
Remove a certain theoretically possible failure case, and force "some"
fallback.
mp_make_wakeup_pipe() always fails on win32. If this call fails on Linux
(and e.g. ao_alsa is used), this will probably burn CPU since poll()
won't work on the invalid file descriptor, but whatever, the failure
case is obscure enough.
There was confusion about what should go into audio pts calculation and
what not (mainly due to the audio push thread). This has been fixed by
using the playing - not written - audio pts (which properly takes into
account the ao's buffer), and incrementing the samples count only by the
amount of samples actually taken from the buffer (unfortunately this
now forces us to keep the lock too long for my taste).
Logic for this was missing from pull.c. For push.c it was missing if the
driver didn't support it. But even if the driver supported it (such as
with ao_alsa), strange behavior was observed by users. See issue #933.
Always check explicitly whether the AO is in paused mode, and if so,
don't drain.
Possibly fixes#933.
CC: @mpv-player/stable
Doesn't work quite right, and will pause for the latency duration after
seeking. Some users use --ao=null to disable audio (even though they
should probably use --no-audio), and this use-case is broken by this
issue too.
CC: @mpv-player/stable
Previous code was completly wrong. This still doesn't report the device
latency, but we report the buffer latency (as before the AO refactoring) and
the AudioUnit's latency (this is a new 'feature').
Apparently we can also report the device actual latency and we should also
calculate the actual sample rate of the audio device instead of using the
nominal sample rate, but I'll leave this for a later commit.
The mplayer1/2/mpv CoreAudio audio output historically contained both usage
of AUHAL APIs (these go through the CoreAudio audio server) and the Device
based APIs (used only for output of compressed formats in exclusive mode).
The latter is a very unwieldy and low level API and pretty much forces us to
write a lot of code for little workr. Also with the widespread of HDMI, the
actual need for outputting compressed audio directly to the device is getting
lower (it was very useful with S/PDIF for bandwidth constraints not allowing
a number if channels transmitted in LPCM).
Considering how invasive it is (uses hog/exclusive mode), the new AO
(`ao_coreaudio_device`) is not going to be autoprobed but the user will have
to select it.
Something like "char *s = ...; isdigit(s[0]);" triggers undefined
behavior, because char can be signed, and thus s[0] can be a negative
value. The is*() functions require unsigned char _or_ EOF. EOF is a
special value outside of unsigned char range, thus the argument to the
is*() functions can't be a char.
This undefined behavior can actually trigger crashes if the
implementation of these functions e.g. uses lookup tables, which are
then indexed with out-of-range values.
Replace all <ctype.h> uses with our own custom mp_is*() functions added
with misc/ctype.h. As a bonus, these functions are locale-independent.
(Although currently, we _require_ C locale for other reasons.)
While I'm not very fond of "const", it's important for declarations
(it decides whether a symbol is emitted in a read-only or read/write
section). Fix all these cases, so we have writeable global data only
when we really need.
So the device buffer can be refilled quickly. Fixes dropouts in certain
cases: if all data is moved from the soft buffer to the audio device
buffer, the waiting code thinks it has to enter the mode in which it
waits for new data from the decoder. This doesn't work, because the
get_space() logic tries to keep the total buffer size down. get_space()
will return 0 (or a very low value) because the device buffer is full,
and the decoder can't refill the soft buffer. But this means if the AO
buffer runs out, the device buffer can't be refilled from the soft
buffer. I guess this mess happened because the code is trying to deal
with both AOs with proper event handling, and AOs with arbitrary
behavior.
Unfortunately this increases latency, as the total buffered audio
becomes larger. There are other ways to fix this again, but not today.
Fixes#818.
Apparently this can happen. So actually only return from waiting if ALSA
excplicitly signals that new output is available, or if we are woken up
externally.
This did not flush remaining audio in the buffer correctly (in case an
AO has an internal block size). So we have to make the audio feed thread
to write the remaining audio, and wait until it's done.
Checking the avoid_ao_wait variable should be enough to be sure that all
data that can be written was written to the AO driver.
This code handles buggy AOs (even if all AOs are bug-free, it's good for
robustness). Move handling of it to the AO feed thread. Now this check
doesn't require magic numbers and does exactly what's it supposed to do.
Until now, we've always calculated a timeout based on a heuristic when
to refill the audio buffers. Allow AOs to do it completely event-based
by providing wait and wakeup callbacks.
This also shuffles around the heuristic used for other AOs, and there is
a minor possibility that behavior slightly changes in real-world cases.
But in general it should be much more robust now.
ao_pulse.c now makes use of event-based waiting. It already did before,
but the code for time-based waiting was also involved. This commit also
removes one awkward artifact of the PulseAudio API out of the generic
code: the callback asking for more data can be reentrant, and thus
requires a separate lock for waiting (or a recursive mutex).
There were subtle and minor race conditions in the pull.c code, and AOs
using it (jack, portaudio, sdl, wasapi). Attempt to remove these.
There was at least a race condition in the ao_reset() implementation:
mp_ring_reset() was called concurrently to the audio callback. While the
ringbuffer uses atomics to allow concurrent access, the reset function
wasn't concurrency-safe (and can't easily be made to).
Fix this by stopping the audio callback before doing a reset. After
that, we can do anything without needing synchronization. The callback
is resumed when resuming playback at a later point.
Don't call driver->pause, and make driver->resume and driver->reset
start/stop the audio callback. In the initial state, the audio callback
must be disabled.
JackAudio of course is different. Maybe there is no way to suspend the
audio callback without "disconnecting" it (what jack_deactivate() would
do), so I'm not trying my luck, and implemented a really bad hack doing
active waiting until we get the audio callback into a state where it
won't interfere. Once the callback goes from AO_STATE_WAIT to NONE, we
can be sure that the callback doesn't access the ringbuffer or anything
else anymore. Since both sched_yield() and pthread_yield() apparently
are not always available, use mp_sleep_us(1) to avoid burning CPU during
active waiting.
The ao_jack.c change also removes a race condition: apparently we didn't
initialize _all_ ao fields before starting the audio callback.
In ao_wasapi.c, I'm not sure whether reset really waits for the audio
callback to return. Kovensky says it's not guaranteed, so disable the
reset callback - for now the behavior of ao_wasapi.c is like with
ao_jack.c, and active waiting is used to deal with the audio callback.
In most places where af_fmt2bits is called to get the bits/sample, the
result is immediately converted to bytes/sample. Avoid this by getting
bytes/sample directly by introducing af_fmt2bps.
In my opinion, we shouldn't use atomics at all, but ok.
This switches the mpv code to use C11 stdatomic.h, and for compilers
that don't support stdatomic.h yet, we emulate the subset used by mpv
using the builtins commonly provided by gcc and clang.
This supersedes an earlier similar attempt by Kovensky. That attempt
unfortunately relied on a big copypasted freebsd header (which also
depended on much more highly compiler-specific functionality, defined
reserved symbols, etc.), so it had to be NIH'ed.
Some issues:
- C11 says default initialization of atomics "produces a valid state",
but it's not sure whether the stored value is really 0. But we rely on
this.
- I'm pretty sure our use of the __atomic... builtins is/was incorrect.
We don't use atomic load/store intrinsics, and access stuff directly.
- Our wrapper actually does stricter typechecking than the stdatomic.h
implementation by gcc 4.9. We make the atomic types incompatible with
normal types by wrapping them into structs. (The FreeBSD wrapper does
the same.)
- I couldn't test on MinGW.
Use the time as returned by mp_time_us() for mpthread_cond_timedwait(),
instead of calculating the struct timespec value based on a timeout.
This (probably) makes it easier to wait for a specific deadline.
This didn't quite work. The main issue was that get_space tries to be
clever to reduce overall buffering, so it will cause the playloop to
decode and queue only as much audio as is needed to refill the AO in
reasonable time. Also, even if ignoring the problem, the logic of the
previous commit was slightly broken. (This required a few retries,
because I couldn't reproduce the issue on my own machine.)
When the audio buffer went low, but could not be refilled yet, it could
happen that the AO playback thread and the decode thread could enter a
wakeup feedback loop, causing up to 100% CPU usage doing nothing. This
happened because the decoder thread would wake up the AO thread when
writing 0 bytes of newly decoded data, and the AO thread in reaction
wakes up the decoder thread after writing 0 bytes to the AO buffer.
Fix this by waking up the decoder thread only if data was actually
played or queued. (This will still cause some redundant wakeups, but
will eventually settle down, reducing CPU usage close to ideal.)
I don't think this is really a very good idea because it is conceptually
incorrect but other prominent multimedia programs use this approach
(VLC and xbmc), and it seems to make the conversion more robust in certain
cases.
For example it has been reported, that configuring a receiver that can output
7.1 to output 5.1, will make CoreAudio report 8 channel descriptions, and the
last 2 descriptions will be tagged kAudioChannelLabel_Unknown.
Fixes#737
The code was falling back to the full waveext chmap_sel when less than 2
channels were detected. This new code is slightly more correct since it only
fills the chmap_sel with the stereo or mono chmap in the fallback case.
CoreAudio supports 3 kinds of layouts: bitmap based, tag based, and speaker
description based (using either channel labels or positional data).
Previously we tried to convert everything to bitmap based channel layouts,
but it turns out description based ones are the most generic and there are
built-in CoreAudio APIs to perform the conversion in this direction.
Moreover description based layouts support waveext extensions (like SDL and
SDR), and are easier to map to mp_chmaps.
The comment says that it wakes up the main thread if 50% has been
played, but in reality the value was 0.74/2 => 37.5%. Correct this. This
probably changes little, because it's a very fuzzy heuristic in the
first place.
Also move down the min_wait calculation to where it's actually used.
For some reason, the buffered_audio variable was used to "cache" the
ao_get_delay() result. But I can't really see any reason why this should
be done, and it just seems to complicate everything.
One reason might be that the value should be checked only if the AO
buffers have been recently filled (as otherwise the delay could go low
and trigger an accidental EOF condition), but this didn't work anyway,
since buffered_audio is set from ao_get_delay() anyway at a later point
if it was unset. And in both cases, the value is used _after_ filling
the audio buffers anyway.
Simplify it. Also, move the audio EOF condition to a separate function.
(Note that ao_eof_reached() probably could/should whether the last
ao_play() call had AOPLAY_FINAL_CHUNK set to avoid accidental EOF on
underflows, but for now let's keep the code equivalent.)
This was reported with PulseAudio 2.1. Apparently it still has problems
with reporting the correct delay. Since ao_pulse.c still has our custom
get_delay implementation, there's a possibility that this is our fault,
but this seems unlikely, because it's full of workarounds for issues
like this. It's also possible that this problem doesn't exist on
PulseAudio 5.0 anymore (I didn't explicitly retest it).
The check is general and works for all push based AOs. For pull based
AOs, this can't happen as pull.c implements all the logic correctly.
This should probably be an AO function, but since the playloop still has
some strange stuff (using the buffered_audio variable instead of calling
ao_get_delay() directly), just leave it and make it more explicit.
Same change as in e2184fcb, but this time for pull based AOs. This is
slightly controversial, because it will make a fast syscall from e.g.
ao_jack. And according to JackAudio developers, syscalls are evil and
will destroy realtime operation. But I don't think this is an issue at
all.
Still avoid locking a mutex. I'm not sure what jackaudio does in the
worst case - but if they set the jackaudio thread (and only this thread)
to realtime, we might run into deadlock situations due to priority
inversion and such. I'm not quite sure whether this can happen, but I'll
readily follow the cargo cult if it makes hack happy.
I'm not quite sure why ao_pulse needs this. It was broken when a thread
to fill audio buffers was added to AO - the pulseaudio callback was
waking up the playback thread, not the audio thread. But nobody noticed,
so it can't be very important. In any case, this change makes it wake up
the audio thread instead (which in turn wakes up the playback thread if
needed).
And also add a function ao_need_data(), which AO drivers can call if
their audio buffer runs low.
This change intends to make it easier for the playback thread: instead
of making the playback thread calculate a timeout at which the audio
buffer should be refilled, make the push.c audio thread wakeup the core
instead.
ao_need_data() is going to be used by ao_pulse, and we need to
workaround a stupid situation with pulseaudio causing a deadlock because
its callback still holds the internal pulseaudio lock.
For AOs that don't call ao_need_data(), the deadline is calculated by
the buffer fill status and latency, as before.
Also fix a format string mistake in a log call using it.
I wonder if this code shouldn't use FormatMessage, but it looks kind
of involved [1], so: no, thanks.
[1] http://support.microsoft.com/kb/256348/en-us
The volume controls in mpv now affect the session's volume (the
application's volume in the mixer). Since we do not request a
non-persistent session, the volume and mute status persist across mpv
invocations and system reboots.
In exclusive mode, WASAPI doesn't have access to a mixer so the endpoint
(sound card)'s master volume is modified instead. Since by definition
mpv is the only thing outputting audio in exclusive mode, this causes no
conflict, and ao_wasapi restores the last user-set volume when it's
uninitialized.
Due to the COM Single-Threaded Apartment model, the thread owning the
objects will still do all the actual method calls (in the form of
message dispatches), but at least this will be COM's problem rather than
having to set up several handles and adding extra code to the event
thread.
Since the event thread still needs to own the WASAPI handles to avoid
waiting on another thread to dispatch the messages, the init and uninit
code still has to run in the thread.
This also removes a broken drain implementation and removes unused
headers from each of the files split from the original ao_wasapi.c.
ao_wasapi.c was almost entirely init code mixed with option code and
occasionally actual audio handling code. Split most things to
ao_wasapi_utils.c and keep the audio handling code in ao_wasapi.c.
Gets rid of the internal ring buffer and get_buffer. Corrects an
implementation error in thread_reset.
There is still a possible race condition on reset, and a few refactors
left to do. If feasible, the thread that handles everything
WASAPI-related will be made to only handle feed events.
Assume obtained.samples contains the number of samples the SDL audio
callback will request at once. Then make sure ao.c will set the buffer
size at least to 3 times that value (or more).
Might help with bad SDL audio backends like ESD, which supposedly uses a
500ms buffer.
In general, we don't need to have a large hw audio buffer size anymore,
because we can quickly fill it from the soft buffer.
Note that this probably doesn't change much anyway. On my system (dmix
enabled), the buffer size is only 170ms, and ALSA won't give more. Even
when using a hardware device the buffer size seems to be limited to
341ms.
This AO pretended to support volume operations when in spdif passthrough
mode, but actually did nothing. This is wrong: at least the GET
operations must write their argument. Signal that volume is unsupported
instead.
This was probably a hack to prevent insertion of volume filters or so,
but it didn't work anyway, while recovering after failed volume filter
insertion does work, so this is not needed at all.
Since the addition of the AO feed thread, 200ms of latency (MIN_BUFFER)
was added to all push-based AOs. This is not so nice, because even AOs
with relatively small buffering (e.g. ao_alsa on my system with ~170ms
of buffer size), the additional latency becomes noticable when e.g.
toggling mute with softvol.
Fix this by trying to keep not only 200ms minimum buffer, but also 200ms
maximum buffer. In other words, never buffer beyond 200ms in total. Do
this by estimating the AO's buffer fill status using get_space and the
initially known AO buffer size (the get_space return value on
initialization, before any audio was played). We limit the maximum
amount of data written to the soft buffer so that soft buffer size and
audio buffer size equal to 200ms (MIN_BUFFER).
To avoid weird problems with weird AOs, we buffer beyond MIN_BUFFER if
the AO's get_space requests more data than that, and as long as the soft
buffer is large enough.
Note that this is just a hack to improve the latency. When the audio
chain gains the ability to refilter data, this won't be needed anymore,
and instead we can introduce some sort of buffer replacement function in
order to update data in the soft buffer.
It is possible to have ao->reset() called between ao->pause() and
ao->resume() when seeking during the pause. If the underlying PCM
supports pausing, resuming an already reset PCM will produce an error.
Avoid that by explicitly checking PCM state before calling
snd_pcm_pause().
Signed-off-by: wm4 <wm4@nowhere>
The uint64_t math would cause overflow at long enough system uptimes
(...such as 3 days), and any precision error given by the double math will
be under one milisecond.
One strange issue is that we apparently can't stop the audio API on
audio reset (ao_driver.reset). We could use SDL_PauseAudio, but that
doesn't specify whether remaining audio is dropped. We also could use
SDL_LockAudio, but holding that over a long time will probably be bad,
and it probably doesn't drop audio. This means we simply play silence
after a reset, instead of stopping the callback completely. (The
existing code ran into an underrun in this situation.)
The delay estimation works about the same. We simply assume that the
callback is locked to audio timing (like ao_jack), and that 1 callback
corresponds to 1 period. It seems this (removed) code fragment assumes
there 1 one period size delay:
// delay subcomponent: remaining audio from the next played buffer, as
// provided by the callback
buffer_interval += callback_interval;
so we explicitly do that too.
Until now, this was always conflated with uninit. This was ugly, and
also many AOs emulated this manually (or just ignored it). Make draining
an explicit operation, so AOs which support it can provide it, and for
all others generic code will emulate it.
For ao_wasapi, we keep it simple and basically disable the internal
draining implementation (maybe it should be restored later).
Tested on Linux only.
Same deal as with the previous commit. We don't lose any functionality,
except for waiting "properly" on audio end, instead of waiting using the
delay estimate.
This removes the ringbuffer management from the code, and uses the
generic code added with the previous commit. The result should be
pretty much the same.
The "estimate" sub-option goes away. This estimation is now always
active. The new code for delay estimation is slightly different, and
follows the claim of the jack framework that callbacks are timed
exactly.
This has 2 goals:
- Ensure that AOs have always enough data, even if the device buffers
are very small.
- Reduce complexity in some AOs, which do their own buffering.
One disadvantage is that performance is slightly reduced due to more
copying.
Implementation-wise, we don't change ao.c much, and instead "redirect"
the driver's callback to an API wrapper in push.c.
Additionally, we add code for dealing with AOs that have a pull API.
These AOs usually do their own buffering (jack, coreaudio, portaudio),
and adding a thread is basically a waste. The code in pull.c manages
a ringbuffer, and allows callback-based AOs to read data directly.
Since the AO will run in a thread, and there's lots of shared state with
encoding, we have to add locking.
One case this doesn't handle correctly are the encode_lavc_available()
calls in ao_lavc.c and vo_lavc.c. They don't do much (and usually only
to protect against doing --ao=lavc with normal playback), and changing
it would be a bit messy. So just leave them.
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.
This field will be moved out of the ao struct. The encoding code was
basically using an invalid way of accessing this field.
Since the AO will be moved into its own thread too and will do its own
buffering, the AO and the playback core might not even agree which
sample a PTS timestamp belongs to. Add some extrapolation code to handle
this case.
Use QueryPerformanceCounter to improve the accuracy of
IAudioClock::GetPosition.
While this is mainly for "realtime correctness" (usually the delay is a
single sample or less), there are cases where IAudioClock::GetPosition
takes a long time to return from its call (though the documentation doesn't
define what a "long time" is), so correcting its value might be important in
case the documented possible delay happens.
The lack of device latency made get_delay report latencies shorter than
they should; on systems with fast enough drivers, the delay is not
perceptible, but high enough invisible delays would cause desyncs.
I'm not yet completely sure whether this is 100% accurate, there are
some issues involved when repeatedly pausing+unpausing (the delay might
jump around by several dozen miliseconds), but seeking seems to be
working correctly now.
The player didn't quit when the end of a file was reached. The reason
for this is that jack reported a constant audio delay even when all
audio was done playing. Whether that was recognized as EOF by the player
depended whether the exact value was higher or lower than the player's
threshhold for what it considers no more audio.
get_delay() should return amount of time it takes until the last sample
written to the audio buffer reaches the speaker. Therefore, we have to
track the estimated time when the last sample is done, and subtract it
from the calculated latency. Basically, the latency is the only amount
of time left in the delay, and it should go towards 0 as audio reaches
ths speakers.
I'm not sure if this is correct, but at least it solves the problem. One
suspicious thing is that we use system time to estimate the end of the
audio time. Maybe using jack_frame_time() would be more correct. But
apart from this, there doesn't seem to be a better way to handle this.
Windows applications that use LoadLibrary are vulnerable to DLL
preloading attacks if a malicious DLL with the same name as a system DLL
is placed in the current directory. mpv had some code to avoid this in
ao_wasapi.c. This commit just moves it to main.c, since there's no
reason it can't be used process-wide.
This change can affect how plugins are loaded in AviSynth, but it
shouldn't be a problem since MPC-HC also does this and it's a very
popular AviSynth client.
1000ms is a bit insane. It makes behavior on playback speed changes
worse (because the player has to catch up the dropped audio due to
audio-chain reset), and perhaps makes seeking slower.
Note that the problem of playback speed changes misbehaving will be
fixed in the future, but even then we don't want to have a buffer that
large.
Always pass around mp_log contexts in the option parser code. This of
course affects all users of this API as well.
In stream.c, pass a mp_null_log, because we can't do it properly yet.
This will be fixed later.
Remove the nonsensical print_lock too.
Things that are called from the option validator are not converted yet,
because the option parser doesn't provide a log context yet.
This could output additional, potentially useful error messages. But the
callback is global and not library-safe, and would require us to add
additional state. Remove it, because it's obviously too much of a pain.
Also, it seems ALSA prints stuff to stderr anyway.
Since m_option.h and options.h are extremely often included, a lot of
files have to be changed.
Moving path.c/h to options/ is a bit questionable, but since this is
mainly about access to config files (which are also handled in
options/), it's probably ok.
The tmsg stuff was for the internal gettext() based translation system,
which nobody ever attempted to use and thus was removed. mp_gtext() and
set_osd_tmsg() were also for this.
mp_dbg was once enabled in debug mode only, but since we have log level
for enabling debug messages, it seems utterly useless.
The previous RING_BUFFER_COUNT value, 64, would have ao_wasapi buffer 64
frames of audio in the ring buffer; a delay of 1280ms, which is clearly
overkill for everything. A value of 8 buffers 8 frames for a total of
160ms.
When get_space was converted to returning samples instead of bytes, a
unit type mismatch in get_delay's calculation returned bogus values. Fix
by converting get_space's value back to bytes.
Fixes playback with ao_wasapi when reaching EOF, or seeking past it.
There are some use cases for this. For example, you can use it to set
defaults of automatically inserted filters (like af_lavrresample). It's
also useful if you have a non-trivial VO configuration, and want to use
--vo to quickly change between the drivers without repeating the whole
configuration in the --vo argument.
This partially reverts commit 7d152965. It turns out that at least some
ALSA drivers (at least snd-hda-intel) report incorrect audio delay with
non-native sample rates, even if the sample rate is only very slightly
different from the native one.
For example, 48000Hz is fine on my hda-intel system, while both 8000Hz
and 47999Hz lead to a delay off by 40ms (according to mpv's A/V
difference display), which suggests that something in ALSA is
calculating the delay using the wrong sample rate.
As an additional problem, with ALSA resampling enabled, using
48001Hz/float/2ch fails, while 49000Hz/float/2ch or 48001Hz/s16/2ch
work. With resampling disabled, all these cases work obviously, because
our own resampler doesn't just refuse any of these formats.
Since some people want to use the ALSA resampler (because it's highly
configurable, supports multiple backends, etc.), we still allow enabling
ALSA resampling with an ao_alsa suboption.
Previous code was using the values of the AudioChannelLabel enum directly to
create the channel bitmap. While this was quite smart it was pretty unreadable
and fragile (what if Apple changes the values of those enums?).
Change it to use a 'dumb' conversion table.
The code stopped at kAudioChannelLabel_TopBackRight and missed mapping for
5 more channel labels. These are in a completely different order that the mpv
ones so they must be mapped manually.
Resampling with non-ancient ALSA setups works fine, so there is no
need to keep this around. Furthermore, as of writing, the default
builtin resampler used by many ALSA setups (taken from libspeex)
actually has higher quality than the default resampling modes of
avresample and swresample.
Simulate proper handling of AOPLAY_FINAL_CHUNK. Print when underruns
occur (i.e. running out of data). Add some options that control
simulated buffer and outburst sizes.
All this is useful for debugging and self-documentation. (Note that
ao_null always was supposed to simulate an ideal AO, which is the reason
why it fools people who try to use it for benchmarking video.)
This should allow it to select better fallback formats, instead of
picking the first encoder sample format if ao->format is not equal to
any of the encoder sample formats.
Not sure what is supposed to happen if the encoder provides no
compatible sample format (or no sample format list at all), but in this
case ao_lavc.c still fails gracefully.
These must be written even if there was no "final frame", e.g. due to
the player being exited with "q".
Although the issue is mostly of theoretical nature, as most audio codecs
don't need the final encoding calls with NULL data. Maybe will be more
relevant in the future.
ao_null should simulate a "perfect" AO, but framestepping behaved quite
badly with it. Framstepping usually exposes problems with AOs dropping
their buffers on pause, and that's what happened here.
Since ao_openal simulates multi-channel audio by placing a bunch of
mono-sources in 3D space, non-interleaved audio is a perfect match for
it. We just have to remove the interleaving code.
ALSA supports non-interleaved audio natively using a separate API
function for writing audio. (Though you have to tell it about this on
initialization.) ALSA doesn't have separate sample formats for this,
so just pretend to negotiate the interleaved format, and assume that
all non-interleaved formats have an interleaved companion format.
Kevin Mitchell
Jonathan Yong
Rudolf Polzer
wm4
Stefano Pigozzi
atomnuker
Amos Onn
Marcoen Hirschberg
Diogo Franco (Kovensky)
foo86
Hans-Kristian Arntzen
James Ross-Gowan
Martin Herkt
bugmen0t
William Light