Basically rewrite all the code supporting the cache (i.e. anything other
than the ringbuffer logic). The underlying design is untouched.
Note that the old cache2.c (on which this code is based) already had a
threading implementation. This was mostly unused on Linux, and had some
problems, such as using shared volatile variables for communication and
uninterruptible timeouts, instead of using locks for synchronization.
This commit does use proper locking, while still retaining the way the
old cache worked. It's basically a big refactor.
Simplify the code too. Since we don't need to copy stream ctrl args
anymore (we're always guaranteed a shared address space now), lots of
annoying code just goes away. Likewise, we don't need to care about
sector sizes. The cache uses the high-level stream API to read from
other streams, and sector sizes are handled transparently.
Before this commit, the cache was franken-hacked on top of the stream
API. You had to use special functions (like cache_stream_fill_buffer()
instead of stream_fill_buffer()), which would access the stream in a
cached manner.
The whole idea about the previous design was that the cache runs in a
thread or in a forked process, while the cache awa functions made sure
the stream instance looked consistent to the user. If you used the
normal functions instead of the special ones while the cache was
running, you were out of luck.
Make it a bit more reasonable by turning the cache into a stream on its
own. This makes it behave exactly like a normal stream. The stream
callbacks call into the original (uncached) stream to do work. No
special cache functions or redirections are needed. The only different
thing about cache streams is that they are created by special functions,
instead of being part of the auto_open_streams[] array.
To make things simpler, remove the threading implementation, which was
messed into the code. The threading code could perhaps be kept, but I
don't really want to have to worry about this special case. A proper
threaded implementation will be added later.
Remove the cache enabling code from stream_radio.c. Since enabling the
cache involves replacing the old stream with a new one, the code as-is
can't be kept. It would be easily possible to enable the cache by
requesting a cache size (which is also much simpler). But nobody uses
stream_radio.c and I can't even test this thing, and the cache is
probably not really important for it either.
The core didn't use these fields, and use of them was inconsistent
accross AOs. Some didn't use them at all. Some only set them; the values
were completely unused by the core. Some made full use of them.
Remove these fields. In places where they are still needed, make them
private AO state.
Remove the --abs option. It set the buffer size for ao_oss and ao_dsound
(being ignored by all other AOs), and was already marked as obsolete. If
it turns out that it's still needed for ao_oss or ao_dsound, their
default buffer sizes could be adjusted, and if even that doesn't help,
AO suboptions could be added in these cases.
Currently every single AO was implementing it's own ringbuffer, many times
with slightly different semantics. This is an attempt to fix the problem.
I stole some good ideas from ao_portaudio's ringbuffer and went from there.
The main difference is this one stores wpos and rpos which are absolute
positions in an "infinite" buffer. To find the actual position for writing /
reading just apply modulo size.
The producer only modifies wpos while the consumer only modifies rpos. This
makes it pretty easy to reason about and make the operations thread safe by
using barriers (thread safety is guaranteed only in the Single-Producer/Single-
Consumer case).
Also adapted ao_coreaudio to use this ringbuffer.
This allows having properties like time-pos in the window title update
properly. There is a danger of this causing significant CPU usage,
depending on the properties used and the window manager.
Instead of implicitly changing the window title on config(), do it as
part of the new VOCTRL.
At first I wanted to make all VOs use the VOCTRL argument directly, but
on a second thought it appears vo_get_window_title() is much more useful
for some (namely, if the window is created lazily on first config()).
Not all VOs are changed. Wayland and OSX have to follow.
%k is not very standard. The manpage notes them as conforming to
"Olson's timezone package", and it's not standard C89, C99 or POSIX.
mingw doesn't provide it, and even some of the smaller Linux libcs
don't have support.
Use %H instead. This gives slightly different results, but I think
this is ok. Difference in behavior between these summarized:
%k: "single digits are preceded by a blank"
%H: "range 00 to 23"
Use the recently introduced screensaver VOCTRLs to control the
screensaver in the X11 backend. This means the behavior when paused
changes: the old code always kept the screensaver disabled, but now the
screensaver is reenabled on pausing.
Rename the --stop-xscreensaver option to --stop-screensaver and make it
more generic. Now it affects all backends that respond to the
screensaver VOCTRLs.
This is slightly better because VOCTRL_RESUME/VOCTRL_PAUSE are usually
needed by VOs to know whether video is actually being played (for
whatever reason), and they wouldn't be passed to the backend's VOCTRL
handler, like vo_x11_control().
Also try to make sure that these flags (both pause state and screensaver
state) are set consistently in some corner cases. For example, it seems
enabling video in the middle of playing a file while the player is
paused would not set the paused flag.
If codec initialization fails, destroy the VO instead of keeping it
around to make sure the state is consistent.
Framestepping is implemented by unpausing the player for the duration of
a frame. Remove the special handling of VOCTRL_PAUSE/RESUME in these
cases. It was most likely needed because these VOCTRLs used to be
important for screen redrawing (blatant guess), which is now handled
completely differently. The only potentially bad side-effect is that the
screensaver will be disabled/reenabled for the duration of one frame.
This workaround sucks. avio just does not support "-" - and ffmpeg's
command line binaries work around it. FOUR TIMES. DIFFERENTLY.
WHY DOESN'T AVIO DO THIS RIGHT TO BEGIN WITH?
Whatever this was supposed to be originally, it doesn't have much value
anymore. It just forced ad_mpg123 to upmix mono to stereo by default
(the audio chain can do that). As an option, it was mostly useless and
misleading, so get rid of it.
Raw MPEG streams can contain PTS discontinuities. While the playback
core has obvious code for handling PTS going backward, PTS going
forward was as far as I can see not handled.
This can be an issue with DVD playback. This wasn't caught earlier,
because DVD playback was just recently switched to demux_lavf, which
implies -no-correct-pts mode. This mode doesn't use PTS in the same way
as the normal playback mode, and as such was too primitive to be
affected by this issue.
Use the following heuristic to handle PTS forward jumps: if the PTS
difference between two frames is higher than 10 seconds, consider it a
reset. (Also, use MSGL_WARN for the PTS discontinuity warnings.)
In this particular case, the MPEG stream was going from pts=304510857
to pts=8589959849 according to ffprobe (raw timestamps), which seems a
bit strange.
Note that this heuristic breaks if the source video has unusually high
frame times. For example Rooster_Teeth_Podcast_191.m4a, an audio file
with a slide show encoded as MJPEG video track.
Some code in mplayer.c did stuff like accessing (dvd_priv_t *)st->priv.
Do this indirectly by introducing STREAM_CTRL_GET_DVD_INFO. This is
extremely specific to DVD, so it's not worth abstracting this further.
This is a preparation for turning the cache into an actual stream, which
simply wraps the cached stream. There are other streams which are
accessed in the way DVD was, at least TV/radio/DVB. We assume these
can't be used with the cache. The code doesn't look thread-safe or fork
aware.
show_chapters, show_tracks, and show_playlist are killed and replaced
with the properties chapter-list, track-list, and playlist. The code
and the output of these stays the same, this is just moving a lot of
code around and reducing the number of properties.
The "old" commands will still be supported for a while (to avoid making
everyone angry), so handle them with the legacy layer. Add something to
suppress printing the legacy warnings for these commands.
Slightly better output when printing ${metadata}. Print each metadata
item as "name: value", instead of the raw list. It's still not very
great, though. The old format is still available through ${=metadata}
for things which dare to use the broken slave mode.
This adds a the property 'clock', which returns the current
local time as the string hh:mm.
Additionally the keybinding 'shift' + 'o' was added to displaying
the clock as '[hh:mm]' .
This commit addresses some issues with the users had with the previous
implementation in commit c39efb9. Here's the changes:
* Use Quartz Event Taps to remove Media Key events mpv handles from
the global OS X queue. This prevents conflicts with iTunes. I did this on
the main thread since it is mostly idling. It's the playloop thread that
actually does all the work so there is no danger of blocking the event tap
callback.
* Introduce `--no-media-keys` switch so that users can disable all of mpv's
media key handling at runtime (some prefer iTunes for example).
* Use mpv's bindings so that users can customize what the media keys do via
input.conf. Current bindings are:
MK_PLAY cycle pause
MK_PREV playlist_prev
MK_NEXT playlist_next
An additional benefit of this implementation is that it is completly handled
by the `macosx_events` file instead of `macosx_application` making the
project organization more straightforward.
This branch heavily refactors the subtitle code (both loading and
rendering), and adds support for a few new formats through FFmpeg.
We don't remove any of the old code yet. There are still some subtleties
related to subreader.c to be resolved: code page detection & conversion,
timing post-processing, UTF-16 subtitle support, support for the -subfps
option. Also, SRT reading and loading ASS via libass should be turned
into proper demuxers. (SRT is needed because Libav's is gravely broken,
and we want ASS loading via libass to cover full libass format support.
Both should be demuxers which are probed _before_ libavformat, so that
all subtitles can be loaded through the demuxer infrastructure, and
libavformat subtitles don't need to be treated in a special way.)
Until now, this happened only when the -no-ass option was used. This
difference in behavior doesn't make much sense, so change it so that
whether -no-ass is used or not doesn't matter. (-no-ass enables the OSD
subtitle renderer, which has the terminal fallback, while the normal
path is video only.)
the changes in set_osd_subtitle() and reinit_video_chain() are for
resetting the state correctly when switching between video/no-video.
Audio and video had their own (very similar) functions to initialize an
AVPacket (ffmpeg's packet struct) from a demux_packet (mplayer's packet
struct). Add a common function for these.
Also use this function for sd_lavc_conv. This is actually a functional
change, as some libavfilter subtitle demuxers add weird out-of-band
stuff as side-data.
Before this, subtitle packets were returned as data ptr/len pairs, and
mplayer.c got the rest (pts and duration) directly from the demuxer
data structures. Then mplayer.c reassembled the packet data structure
again.
Pass packets directly instead. The mplayer.c side stays a bit awkward,
because the (now by default unused) DVD path keeps getting in the way.
In demux.c there's lots of weird stuff (3 functions that read packets,
really?), but we want to keep the code equivalent for now to avoid
hitting weird issues and corner cases.
The -no-ass option used to disable all use of libass completely. This
doesn't work this way anymore, and the text subtitle path has an
inherent dependency on libass. Currently -no-ass does 3 things:
1. Strip tags and formatting on display, and use a separate renderer for
the result. (Which might be the terminal, or libass via OSD code.)
2. Not loading attached fonts from Matroska files.
3. Use subreader.c instead of libass for reading .ass files.
1. and 2. are ok and what the user (probably wants), but 3. doesn't
really make sense anymore. subreader.c reads .ass files just fine, but
then does some strange things to them (something about coalescing and
re-adding newlines?), leading to even more broken display with -no-ass.
Instead of fighting with subreader.c, just use libass as loader.
This means subassconvert.c is split in sd_srt.c and sd_microdvd.c. Now
this code is involved in the sub conversion chain like sd_movtext is.
The invocation of the converter in sd_ass.c is removed.
This requires some other changes to make the new sub converter code work
with loading external subtitles. Until now, subtitles loaded via
subreader.c was assumed to be in plaintext, or for some formats, in ASS
(except in -no-ass mode). Then these were added to an ASS_Track. Change
this so that subtitles are always in their original format (as far as
decoders/converters for them are available), and turn every sub event
read by subreader.c as packet to the dec_sub.c subtitle chain.
This removes differences between external/demuxed and -ass/-no-ass code
paths further.
After killing the non functional AR support in c8fd9e5 I got much complaints so
this adds AR support back in (and it works). I am using the HIDRemote class by
Felix Schwarz and that part of the code is under the BSD license. I slightly
modified it replacing [NSApplication sharedApplication] with NSApp. The code
of the class is quite complex (probably because it had to deal with all the
edge cases with IOKit) but it works nicely as a black box.
In a later commit I'll remove the deprecation warnings caused by HIDRemote's
usage of Gestalt.
Check out `etc/input.conf` for the default bindings.
Apple Remote functionality is automatically compiled in when cocoa is enabled.
It can be disabled at runtime with the `--no-ar` option.
On OSX with Cocoa enabled keyDown events are now handled with
addLocalMonitorForEventsMatchingMask:handler:. This allows to respond to
events even when there is no VO initialized but the GUI is focused.
Add a basic infrastructure for subtitle converters. These converters
work sort-of like decoders, except that they produce packets instead
of subtitle bitmaps. They are put in front of actual decoders.
Start with sd_movtext. 4 lines of code are blown up to a 55 lines file,
but fortunately this is not going to be that bad for the following
converters.
OPT_STRING_VALIDATE actually did nothing. This made -vo opengl crash or
misbehave when passing an invalid value for the lscale, cscale or 3dlut-
size (the only users for this option type).
The code added with this commit was either blatantly forgotten with the
commit introducing this option type, or somehow lost.
Make the sub decoder stuff independent from sh_sub (except for
initialization of course). Sub decoders now access a struct sd only,
instead of getting access to sh_sub. The glue code in dec_sub.c is
similarily independent from osd.
Some simplifications are made. For example, the switch_id stuff is
unneeded: the frontend code just has to make sure to call osd_changed()
any time subtitles are switched.
This is also preparation for introducing subtitle converters. It's much
cleaner to completely separate demuxer header/renderer glue/decoders
for this purpose, especially since sub converters might completely
change how demuxer headers have to be interpreted.
Also pass data as demux_packets. Currently, this doesn't help much, but
libavcodec converters might need scary stuff like packet side data, so
it's perhaps better to go with passing packets.
Subtitle files are opened in mplayer.c, not using the demuxer
infrastructure in general. Pretend that this is not the case (outside of
the loading code) by opening a pseudo demuxer that does nothing. One
advantage is that the initialization code is now the same, and there's
no confusion about what the difference between track->stream,
track->sh_sub and mpctx->sh_sub is supposed to be.
This is a bit stupid, and it would be much better if there were proper
subtitle demuxers (there are many in recent FFmpeg, but not Libav). So
for now this is just a transition to a more proper architecture. Look
at demux_sub like an artifical limb: it's ugly, but don't hate it - it
helps you to get on with your life.
This unifies the subtitle rendering path. Now all subtitle rendering
goes through sd_ass.c/sd_lavc.c/sd_spu.c.
Before that commit, the spudec.h functions were used directly in
mplayer.c, which introduced many special cases. Add sd_spu.c, which is
just a small wrapper connecting the new subtitle render API with the
dusty old vobsub decoder in spudec.c.
One detail that changes is that we always pass the palette as extra
data, instead of passing the libdvdread palette as pointer to spudec
directly. This is a bit roundabout, but actually makes the code simpler
and more elegant: the difference between DVD and non-DVD dvdsubs is
reduced.
Ideally, we would just delete spudec.c and use libavcodec's DVD sub
decoder. However, DVD playback with demux_mpg produces packets
incompatible to lavc. There are incompatibilities the other way around
as well: packets from libavformat's vobsub demuxer are incompatible to
spudec.c. So we define a new subtitle codec name for demux_mpg subs,
"dvd_subtitle_mpg", which only sd_spu can decode.
There is actually code in spudec.c to "assemble" fragments into complete
packets, but using the whole spudec.c is easier than trying to move this
code into demux_mpg to fix subtitle packets.
As additional complication, Libav 9.x can't decode DVD subs correctly,
so use sd_spu in that case as well.
It appears demux_mpg doesn't output timestamps for subtitles. The vobsub
code handled this by doing its own PTS calculations. This code is absent
from the normal subtitle decoder path. Copy this code into the normal
path, so that we can unify the subtitle decoder paths in a later commit.
Decoding subtitles with sd_lavc when playing DVD with demux_mpg still
doesn't work.
The -no-ass switch used to disable any use of libass for text subtitles.
This is not really the case anymore, because libass is now always
involved when rendering text. The only remaining use of -no-ass is
disabling styling or showing subtitles on the terminal. On the other
hand, the old subtitle rendering path is a big reason why the subtitle
code is still a big mess with an awful number of obscure special cases.
In order to simplify it, remove the old subtitle rendering code, and
always go through sd_ass.c. Basically, we use ASS_Track as central data
structure for storing text subtitles instead of struct sub_data. This
also makes libass mandatory for all text subs, even if they are printed
to the terminal in -no-video mode. (We could add something like sd_text
to avoid this, but it's not worth the trouble.)
struct sub_data and subreader.c are still around, even its ASS/SSA
reader. But struct sub_data is freed right after converting it to
ASS_Track. The internal ASS reader actually can handle some obscure
cases libass can't, like files encoded in UTF-16.
The core deselected all streams on initialization, and then selected the
streams it actually wanted. This was no problem for
demux_mkv/demux_lavf, but old demuxers (like demux_asf) could lose some
packets. The problem is that these demuxers can buffer some data on
initialization, which then is flushed on track switching. Fix this by
explicitly avoiding deselecting a wanted stream.
Most of these are rather questionable, the rest you rarely need to set
manually. You still can set all of them with -lavdopts-o (because
libavcodec has AVOptions for them).
Playing something with "mpv f1.mkv f2.mkv --gapless-audio --volume=20"
caused the volume to be reset when playing a new file. Normally, the
volume should not be reset (unless explicitly requested with per-file
options), and without either --gapless-audio or --volume it works as
expected.
The underlying problem is that volume was saved only when the AO was
uninitialized, and also the volume was always set when starting a file.
Fix this by saving the volume when playback ends, and when the audio
is reinitialized. To make sure the volume is never restored twice or
saved in the wrong situation, introduce INITIALIZED_VOL.
Also note that this volume saving and restoring only happens if the
--volume option is used. mixer.c does its own bookkeeping of volume.
The main reason for this is that the volume option could be reset by
per-file options (see manpage), and mixer.c doesn't know anything
about this stuff. This is probably dumb, and maybe some things could
be simplified. But for now this will work.
When AAC is streamed over HTTP, using libavformat defaults is
pathetically slow. One solution for that is skipping probing and using
the mimetype to identify that it's AAC instead. This is what we did
before this commit (and ffmpeg does it too, but their logic is too
"inaccessible" for mpv).
This is still pretty fragile though. Make it a bit more robust by
requiring minimal probing. A probescore of 25 is reached after feeding
2 KB to libavformat (instead of > 500 KB for the normal probescore), so
use that. This is done only when streaming AAC from HTTP to reduce the
possibility of weird breakages for other formats.
Also reduce analyzeduration. The default analyzeduration will make
libavformat read lots of data, which makes playback start slow. So we
set analyzeduration to a low value. On the other hand, doing that for
other formats is risky, because there are unspecified effects with
certain "strange" formats (like transport streams). So we do this only
if we're streaming AAC from HTTP as well.
tl;dr libavformat is shit for media players
This can control whether demux_lavf should use the HTTP mime type to
determine the format, instead of probing the data with the libavformat
API. Do this to allow easier debugging in case the mimetype is
incorrect. (This is done only for AAC streams right now.)
In commit 0e07189, I made the status line always print a newline,
instead of cutting the output at 80 columns (or if stderr is a terminal,
whatever width the terminal reports). This is better in the case the
output goes into a log file or a pipe.
This caused problems for people who want to pipe raw video to mpv, so
change it again. (Not sure why they won't use FIFOs instead.)
Now output untrimmed lines if the slave mode flag is set, which makes
sense to do, too. The current slave mode is still on life support,
though.
GetTimer() is generally replaced with mp_time_us(). Both calls return
microseconds, but the latter uses int64_t, us defined to never wrap,
and never returns 0 or negative values.
GetTimerMS() has no direct replacement. Instead the other functions are
used.
For some code, switch to mp_time_sec(), which returns the time as double
float value in seconds. The returned time is offset to program start
time, so there is enough precision left to deliver microsecond
resolution for at least 100 years. Unless it's casted to a float
(or the CPU reduces precision), which is why we still use mp_time_us()
out of paranoia in places where precision is clearly needed.
Always switch to the correct time. The whole point of the new timer
calls is that they don't wrap, and storing microseconds in unsigned int
variables would negate this.
In some cases, remove wrap-around handling for time values.
This was used by some VOs to do timing of cursor autohiding, but we
recently moved that out of the VOs. Even though this mechanism might
be a good idea and could be needed again in future (but for what?),
it's unused now. So better just get rid of it.
Make OS specific timer code export a mp_raw_time_us() function, and
add generic implementations of GetTimer()/GetTimerMS() using this
function. New mpv code is supposed to call mp_time_us() in situations
where precision is absolutely needed, or mp_time_s() otherwise.
Make it so that mp_time_us() will return a value near program start.
We don't set it to 0 though to avoid confusion with relative vs.
absolute time. Instead, pick an arbitrary offset.
Move the test program in timer-darwin.c to timer.c, and modify it to
work with the generic timer functions.
If VO deinterlacing is unavailable, try to insert vf_yadif.
If vf_lavfi is available, actually use vf_yadif from libavfilter. The
libavfilter version of this filter is faster, more correct, etc., so it
is preferred. Unfortunately vf_yadif obviously doesn't support
VFCTRL_GET/SET_DEINTERLACE, and with the current state of the
libavfilter API, it doesn't look like there is any simple way to
emulate it. Instead, we simply insert the filter with a specific label,
and if deinterlacing is to be disabled, the filter is removed again by
label.
This won't do the right thing if the user inserts any deinterlacing
filter manually (except native vf_yadif, which understands the VFCTRL).
For example, with '-vf lavfi=yadif', pressing 'D' (toggle deinterlacing)
will just insert a second deinterlacer filter. In these cases, the user
is supposed to map a command that toggles his own filter instead of
using 'D' and the deinterlace property.
The same applies if the user wants to pass different parameters to the
deinterlacer filters.
If a complete filter description is passed to -vf-del, search for an
existing filter with the same label or the same name/arguments, and
delete it. The rules for filter entry equality are the same as with
the -vf-toggle option.
E.g.
-vf-add gradfun=123:gradfun=456
-vf-del gradfun=456
does what you would expect.
Can be used to refer to filters by name. Intended to be used when the
filter chain is changed at runtime.
A label can be assigned to a filter by prefixing it with '@name:', where
'name' is an user-chosen identifier. For example, a filter added with
'-vf-add @label1:gradfun=123' can be removed with '-vf-del @label1'.
If a filter with an already existing label is added, the existing filter
is replaced with the new filter (this happens for both -vf-add and
-vf-pre). If a filter is replaced, the new filter takes the position of
the old filter, instead of being appended/prepended to the filter chain
as usual. For -vf-toggle, labels are compared if at least one of the
filters has a label; otherwise they are compared by filter name and
arguments (like before). This means two filters are never considered
equal if one has a label and the other one does not.
This prefers ./ on Windows if-and-only-if the file being searched for
already exists there. (If the mpv directory is non-writable, the result
is still intended behavior.) This change is transparent to most users
because the user has to move the config files there intentionally, and
if anything, not being detected would be the surprising behavior.
In the long run this should be done differently. ID_... output sucks.
This commit will be reverted as soon as I have a good idea how this
should be done properly.
The vf-toggle option parsing (normally used for runtime video filter
switching only) was missing comparing the parameter values. Fix this,
and also make the code a bit more robust.
Also add a "raw" prefix for commands, which prevents property expansion.
The idea is that if the commands are generated by a program, it doesn't
have to know whether the command expands properties or not.
This is more consistent, and doesn't bother the user with ordering
rules when new prefixes are added.
Will break obscure uses of legacy commands: if the command is supposed
to be translated by the legacy command bridge, and if that command uses
one of the pausing* prefixes, the command can't be parsed. Well, just
use the new commands in this case.
Add the "vf" command, which allows changing the video filter chain at
runtime. For example, the 'y' key could be bound to toggle deinterlacing
by adding 'y vf toggle yadif' to the input.conf.
Reconfiguring the video filter chain normally resets the VO, so that it
will be "stuck" until a new video frame is rendered. To mitigate this, a
seek to the current position is issued when the filter chain is changed.
This is done only if playback is paused, because normal playback will
show an actual new frame quickly enough.
If vdpau hardware decoding is used, filter insertion (whether it fails
or not) will break the video for a while. This is because vo_vdpau
resets decoding related things on vo_config().
With the current semantics, there's no reason to disallow this.
(Although in my opinion, -vf should rather map to -vf-add than -vf-set,
however that is an independent issue from this change.)
Works like -vf-add, except if a filter already exists and has the same
parameters, it's removed instead of added.
Not really useful on the command line itself, but will make sense for
runtime filter changing in the following commit.
Until now, -vf-del required a list of indexes. This was a bit
inconvenient, so add support for using filter names too. Also simplify
the code a bit, doing the change would have been too painful otherwise.
The main() function is special, and omitting the return statement would
make it always return 0. And also, mpv_main() actually never returns, it
calls exit() through exit_player() instead. But change it anyway,
because it looks misleading.
Apparently useful for dumping DVD. Could also be used to rip streams
with libquvi and such, but for that there are better tools. Actually
I doubt there aren't better tools to dump DVDs, but whatever, this was
a feature request, so I don't need a good reason.
This helps passing the channel layout correctly from decoder to audio
filter chain. (Because that part "reuses" the demuxer level codec
parameters, which is very disgusting.)
Note that ffmpeg stuff already passed the channel layout via
mp_copy_lav_codec_headers(). So other than easier dealing with the
demuxer/decoder parameters mess, there's no real advantage to doing
this.
Make the --channels option accept a channel map. Since simple numbers
map to standard layouts with the given number of channels, this is
downwards compatible. Likewise for demux_rawaudio.
This actually breaks audio for 5/6/8 channels. There's no reordering
done yet. The actual reordering will be done inside of af_lavrresample
and has to be made part of the format negotiation.
This adds Mission Control fullscreen functionality to mpv. Since this doesn't
play well with many of mpv's features disable it by default. Users can activate
this feature by using `--native-fs` when starting mpv.
Fixes#34
This commit is a followup on the previous one and uses a solution I like more
since it totally decouples the Cocoa code from mpv's core and tries to emulate
a generic Cocoa application's lifecycle as much as possible without fighting
the framework.
mpv's main is executed in a pthread while the main thread runs the native cocoa
event loop.
All of the thread safety is mainly accomplished with additional logic in
cocoa_common as to not increase complexity on the crossplatform parts of the
code.
Schedule mpv's playloop as a high frequency timer inside the main Cocoa event
loop. This has the benefit to allow accessing menus as well as resizing the
window without the playback being blocked and allows to remove countless hacks
from the code that involved manually pumping the event loop as well simulating
manually some of the Cocoa default behaviours.
A huge improvement consists in removing NSApplicationLoad. This is a C function
defined in the Cocoa header and implements a minimal OSX application under ther
hood so that you can use the Cocoa GUI toolkit from C/C++ without having to
respect the Cocoa standards in terms of application initialization. This was
bad because the behaviour implemented by NSApplicationLoad was hard to customize
and had several gotchas especially in the menu department.
mpv was changed to be just a nib-less application. All the Cocoa part is still
generated in code but the event handling is now not dissimilar to what is
present in a stock Mac application.
As a part of reviewing the initialization process, I also removed all of
`osdep/macosx_finder_args`. The useful parts of the code were moved to
`osdep/macosx_appication` which has the broaded responsibility of managing the
full lifecycle of the Cocoa application. By consequence the
`--enable-macosx-finder` configure switch was killed as well, as this feature
is always enabled.
Another change the users will notice is that when using a bundle the `--quiet`
option will be inserted much earlier in the initializaion process. This results
in mpv not spamming mpv.log anymore with all the initialization outputs.
Makes it easier to understand... maybe. It's still pretty strange how
this function may either queue the seek or seek immediately. The way
it actually works doesn't change, queuing the seek is just moved into
the function.
Also add a execute_queued_seek() function, which resets the queue state
correctly.
The frontend doesn't use this.
Also use double for returning the chapter times. Everything uses double
for times, and there's no reason to use float here.
These were found by the cppcheck and scan-build static analyzers. Most
of these aren't interesting (the 2 previous commits fix some interesting
cases found by these analyzers), and they don't nearly fix all warnings.
(Most of the unfixed warnings are spam, things MPlayer never cared
about, or false positives.)
A "watch later" command is now mapped to Shift+Q. This quits the player
and stores the playback state in a config file in ~/.mpv/watch_later/.
When calling the player with the same file again, playback is resumed
at that time position.
It's also possible to make mpv save playback state always on quit with
the --save-position-on-quit option. Likewise, resuming can be disabled
with the --no-resume-playback option.
This also attempts to save some playback parameters, like fullscreen
state or track selection. This will unconditionally override config
settings and command line options (which is probably not what you would
expect, but in general nobody will really care about this). Some things
are not backed up, because that would cause various problems. Additional
subtitle files, video filters, etc. are not stored because that would be
too hard and fragile. Volume/mute state are not stored because it would
mess up if the system mixer is used, or if the system mixer was
readjusted in the meantime.
Basically, the tradeoff between perfect state restoration and
complexity/fragility makes it not worth to attempt to implement
it perfectly, even if the result is a little bit inconsistent.
Now vid/aid/sid can be used as properties. video/audio/sub still work,
but they are aliases for the "real" properties.
This guarantees that options/properties use the same value range. One
consequence is that the video/audio/sub properties return "no" as value
if no track is selected instead of -1.
Also, mark demuxer as not capable if DVD playback is done. The problem
with DVD is that playback time (stream_pts) is not reported frame-exact,
and the time is a "guess" at best.
YCgCo can be manually selected, but will also be used if the decoder
reports YCgCo. To make things more fun, files are sometimes marked
incorrectly, which will display such broken files incorrectly starting
with this commit.
This is an attempt to make quoting of sub-option values less awkward,
even if it works only with some shells. This is needed mainly for
vf_lavfi. Also update the vf_lavfi manpage section.
Rename the struct MPOpts "start_pause" field to "pause". Store the user-
pause state in that field, so that both runtime pause toggling and the
--pause switch change the same variable. Simplify the initialization of
pause so that using --pause and changing the file while paused is
exactly the same case (changing the file while paused doesn't unpause,
this has been always this way).
Also make it a bit more consistent. Before, starting with --pause would
reset the pause state for every file, instead of following the usual
semantics for option switches (compare with behavior of --fs).
The core pauses and unpauses automatically to wait for the network
cache (also known as buffering). This conflicted with user pause
control, and was perceived as if the player was unresponsive and/or
the cache just overturned the user's decisions.
Change it so that the actual pause state and the pause state as
intended by the user never conflict. If the user toggles pause, the
pause state will be in the expected state as soon as the cache is
loaded.
There were two problems.
First, frames past the end of the current segment were added to the
index, which messed up backstepping. Check for the endpts before
added a frame to the index.
Second, it wasn't possible to step over segments which change the file.
Changing a file causes decoder reinitialization, which (rightfully)
is treated as discontinuity (and vo_pts_history_seek_ts was changed).
Add some extra code to pretend that a segment-switching seek/reinit
does not introduce discontinuities.
There's still a weird corner case: sometimes, you can frame step forward
on the last frame of a segment without reaching the next segment
immediately. This is because the playloop switches into audio-only mode.
The segment is switched when both audio and video have ended, so the
frame stepping will play random sized chunks of audio until the segment
will be switched. This gives the impression that backstepping doesn't
work perfectly, even though it's the other way around and frame stepping
behaves weird. This is a consequence of wanting to make frame stepping
work with audio, and is not really a bug.
Allows stepping back one frame via the frame_back_step inout command,
bound to "," by default.
This uses the precise seeking facility, and a perfect frame index built
on the fly. The index is built during playback and precise seeking, and
contains (as of this commit) the last 100 displayed or skipped frames.
This index is used to find the PTS of the previous frame, which is then
used as target for a precise seek. If no PTS is found, the core attempts
to do a seek before the current frame, and skip decoded frames until the
current frame is reached; this will create a sufficient index and the
normal backstep algorithm can be applied.
This can be rather slow. The worst case for backstepping is about the
same as the worst case for precise seeking if the previous frame can be
deduced from the index. If not, the worst case will be twice as slow.
There's also some minor danger that the index is incorrect in case
framedropping is involved. For framedropping due to --framedrop, this
problem is ignored (use of --framedrop is discouraged anyway). For
framedropping during precise seeking (done to make it faster), we try
to not add frames to the index that are produced when this can happen.
I'm not sure how well that works (or if the logic is sane), and it's
sure to break with some video filters. In the worst case, backstepping
might silently skip frames if you backstep after a user-initiated
precise seek. (Precise seeks to do indexing are not affected.)
Likewise, video filters that somehow change timing of frames and do not
do this in a deterministic way (i.e. if you seek to a position, frames
with different timings are produced than when the position is reached
during normal playback) will make backstepping silently jump to the
wrong frame. Enabling/disabling filters during playback (like for
example deinterlacing) will have similar bad effects.
It's not sure if there's anything that could trigger this accidentally.
Normally this can't happen, because hrseek ends always if the PTS is
large enough, the same condition which disables framedrop. Seeking
resets hrseek framedrop anyway.
On the other hand, this change makes the code easier to understand,
and might be more robust against weird corner cases.
Block X11's native key repeat, and use mpv's key repeat handling in
input.c instead.
No configure check for XKB. Even though it's an extension, it has been
part of most (all?) xlibs since 1996. If XKB appears to be missing,
just refuse enabling x11.
This is a potentially controversial change. mpv will use its own key
repeat rate, instead of X11's. This should be better, because seeking
will have a standardized "speed" (seek events per seconds when keeping
a seek key held down). It will also allow disabling key repears for
certain commands, though this is not done anywhere yet.
The new behavior can be disabled with the --native-keyrepeat option.
Key repeats were skipped when playloop iterations took too long. Fix
this by using the total times for key repeat calculation, instead of the
time difference to the last key repeat event.
Key bindings can include mutiple keys at once (additional to key
modifiers like ctrl etc.). This becomes annoying when quickly switching
between two bound keys, e.g. when seeking back and forth, you might end
up hitting the "left" and "right" keys at once. The user doesn't expect
to invoke the key binding "left-right", but would prefer a key stroke to
invoke the binding it was supposed to invoke.
So if there's no binding for a multi-key combination, try to find a
binding for the key last held down. This preserves the ability to define
multi-key combinations, while the common case works as expected.
VOs can use the MP_KEY_STATE_DOWN modifier to pass key up/down events to
input.c, instead of just simple key presses. This allows doing key auto-
repeat handling in input.c, if the VO doesn't want to do that.
One issue is that so far, this code has been used only for mouse events,
even though the code was originally written with keyboard keys in mind.
One difference between mouse keys and keyboard keys is that the initial
key down should not generate an input command with mouse buttons
(input.c did that), while keyboard events should (input.c didn't do
that). Likewise, releasing a key should generate input commands for
mouse buttons releases, but not for the keyboard.
Change the code so mouse buttons (recognized via the MP_NO_REPEAT_KEY
flag) follow the old hehavior, while other keys generate input commands
on key down, but not on key release.
Note that a key release event is posted either using
MP_INPUT_RELEASE_ALL, or a normal key press event after having sent a an
event with MP_KEY_STATE_DOWN. This is probably a bit confusing, and a
MP_KEY_STATE_RELEASE should be added.
Fix shift-handling with MP_KEY_STATE_DOWN as well.
Empty sub-option parameters mean the sub-option should be skipped,
e.g. -vf gradfun=:10 sets the second option (by position) to 10. This
was broken in commit 04f1e2d.
This allows things like:
'--vf=lavfi="gradfun=20:30"'
Adjust the documentation for vf_lavfi to make the example less verbose.
As an unrelated change, add a general description to vf_lavfi.
Requires recent FFmpeg/Libav git versions. Earlier versions will not
be supported, as the API is different. (A libavfilter version that
uses AVFrame instead of AVFilterBuffer is needed.)
Note that this is sort of useless, because the option parser prevents
you from making use of the full libavfilter graph syntax. This has to be
fixed later.
Most of the filter creation code (half of the config() function) has
been taken from avplay.c.
This code is not based on MPlayer's vf_lavfi. The MPlayer code doesn't
compile as it hasn't been updated through multiple libavfilter API
changes, making it completely useless as a starting point.
Parsing sub-configs (like --rawvideo=subopts or the suboptions for
--vo=opengl:subopts) was completely different from the -vf parsing code
for a variety of reasons. This change at least makes -vf use the same
splitter code as sub-config options.
The main improvement is that -vf now accepts quotes, so you can write
things like:
-vf 'lavfi=graph="gradfun=10:20"'
(The '' quotes are for shell escaping.)
This is a rather big and intrusive change. Trying some -vf lines from
etc/encoding-example-profiles.conf seems to confirm it still works.
This also attempts to unify one subtle difference in handling of
positional arguments. One consequence is that a minor detail changes.
Sub-configs don't know positional arguments, and something like "--
opt=sub1=val1:sub2" means that sub2 has to be a flag option. In -vf
parsing, sub2 would be a positional option value. To remove this
conflict and to facilitate actual unification of the parsers in the
future, the sub2 will be considered a flag option if and only if such a
flag option exists. Otherwise, it's considered a value for a positional
option.
E.g. if there's a filter "foo" with a string option "sopt" and a flag
option "fopt", the behavior of the following changes:
-vf foo=fopt
Before this commit, this would set "sopt=fopt" in the filter. Now, it
enables the fopt flag, and the sopt option remains unset. This is not an
actual problem to my knowledge.
Remove the "object settings" based track range parsing (needed by
stream_cdda only), and make stream_cdda use CONF_TYPE_INT_PAIR.
This makes the -vf parsing code completely independent from other
options. A bit of that code was used by the mechanism removed with
this commit.
Before this commit, it was more or less random which subtitle was
preferred if there was both an auto-loaded external subtitle, and a
subtitle loaded via -sub or -subfile. -sub subtitles happened to be
preferred over auto-loaded subs, while -subfile didn't. Fix the -subfile
case, and make the behavior consistent by making the selection behavior
explicit.
Get rid of the 1-char subtitle type field. Use sh_stream->codec instead
just like audio and video do. Use codec names as defined by libavcodec
for simplicity, even if they're somewhat verbose and annoying.
Note that ffmpeg might switch to "ass" as codec name for ASS, so we
don't bother with the current silly "ssa" name.
Matroska files can contain multiple segments, which are literally
further Matroska files appended to the main file. They can be referenced
by segment linking.
While this is an extraordinarily useless and dumb feature, we support it
for the hell of it.
This is implemented by adding a further demuxer parameter for skipping
segments. When scanning for linked segments, each file is opened
multiple times, until there are no further segments found. Each segment
will have a separate demuxer instance (with a separate file handle
etc.).
It appears the Matroska spec. has an even worse feature for segments:
live streaming can completely reconfigure the stream by starting a new
segment. We won't add support for it, because there are 0 people on this
earth who think Matroska life streaming is a good idea. (As opposed to
serving Matroska/WebM files via HTTP.)
With Matroska ordered chapters, the main file (i.e. the file you're
playing) can be empty, while all video/audio data is in linked files.
Some files don't even contain the track list, only chapter information.
mpv refused to play these, because normally, the main file dictates the
track layout.
Fix this by using the first segment for track data if no part of the
timeline is sourced from the main file.
I am aware this detection may occur too late, depending on other
settings, but at least it usually works and is portable.
Where the output fd can be changed, though, it'd be better to force a
similar behaviour via file descriptor use: use pipe:3 as output to FD 3,
and change the calling program to expect the stream on FD 3.
Anything this option did has been removed in the preceding 3 commits.
Note that even though these options sounded like a good idea (like
setting accuracy vs. speed tradeoffs), they were not really properly
implemented.
Switch the internal channel order to libavcodec's. If the channel number
mismatches at some point, use libavresample for up- or downmixing.
Remove the old af_pan automatic downmixing.
The libavcodec channel order should be equivalent to WAVEFORMATEX order,
at least nowadays. reorder_ch.h assumes that WAVEFORMATEX and libavcodec
might be different, but all defined channels have the same mappings.
Remove the downmixing with af_pan as well as the channel conversion with
af_channels from af.c, and prefer af_lavrresample for this. The
automatic downmixing behavior should be the same as before (if the
--channels option is set to 2, which is the default, the audio output
is forced to 2 channels, and libavresample does all downmixing).
Note that mpv still can't do channel layouts. It will pick the default
channel layout according to the channel count. This will be fixed later
by passing down the channel layout as well.
af_hrtf depends on the order of the input channels, so reorder to ALSA
(for which this code was written). This is better than changing the
filter code, which is more risky.
ao_pulse can accept waveext order directly, so set that as channel
mapping.
Make sure automatically inserted filters are removed on full reinit
(they are re-added later if they are really needed). Automatically
inserted filters were never explicitly removed, instead, it was
expected that redundant conversion filters detach themselves. This
didn't work if there were several chained format conversion filters,
e.g. s16le->floatle->s16le, which could result from repeated filter
insertion and removal. (format filters detach only if input format and
output format are the same.)
Further, the dummy filter (which exists only because af.c can't handle
an empty filter chain for some reason) could introduce bad conversions
due to how the format negotiation works. Change the code so that the
dummy filter never takes part on format negotiation. (It would be better
to fix format negotiation, but that would be much more complicated and
would involving fixing all filters.)
Simplify af_reinit() and remove the start audio filter parameter. This
means format negotiation and filter initialization is run more often,
but should be harmless.
This option can be used to selectively reset settings when playing the
next file in the playlist (i.e. restore mplayer and mplayer2 behavior).
Might remove this option again should it turn out that nobody uses it.
Consider:
mpv --volume 10 file1.mkv file2.mkv
Before this commit, the volume was reset to 10 when playing file2.mkv.
This was inconsistent to most other options. E.g. --brightness is a
rather similar case.
In general, settings should never be reset when playing the next file,
unless the option was explicitly marked file-local. This commit
corrects the behavior of the --volume and --mute options.
File local --volume still works as expected:
mpv --{ --volume 10 file1.mkv file2.mkv --}
This sets the volume always to 10 on playback start.
Move the m_config_leave_file_local() call down so that the mixer code
in uninit_player() can set the option volume and mute variables without
overwriting the global option values.
Another subtle issue is that we don't want to set volume if there's no
need to, which is why the user_set_volume/mute fields are introduced.
This is important because setting the volume might change the system
volume depending on other options.
Commit bc20f2c moved the variable default_max_pts_correction (which
backs the --mc option) to the MPOpts struct. The initializer was
forgotten when doing this, so it was left at 0. This disabled part of
the A/V sync mechanism. This was apparent when using ad_spdif (this
decoder has other A/V sync related problems, and thus triggers this
issue easily).
Closes#59.
Makes sure that seeking to a given time position shows the subtitle at
that position. This can fail if the subtitle packet is not close enough
to the seek target. Always enabled for hr-seeks, and can be manually
enabled for normal seeks with --mkv-subtitle-preroll.
This helps displaying subtitles correctly with ordered chapters. When
switching ordered chapter segments, a seek is performed. If the subtitle
is timed slightly before the start of the segment, it normally won't be
demuxed. This is a problem with all seeks, but in this case normal
playback is affected. Since switching segments always uses hr-seeks,
the code added by this commit is always active in this situation.
If no subtitles are selected or the subtitles come from an external
file, the demuxer should behave exactly as before this commit.
In the last cleanup round, this was accidentally changed from a store
option to int, and the option value was passed as flag value.
(Not that anyone needs/uses/cares about this option...)
The main problem with video PTS was that it wasn't very useful when
playing audio files with cover art. Using the audio time instead was an
obvious solution. Unfortunately, this leads to "inexact" reporting of
the playback time in paused mode, and audio is always ahead by small,
essentially random amounts of time ahead. This is possibly because the
times reported by AOs are not entirely accurate when paused (see commit
9b3bf76).
Switch back to video PTS, and use a simpler way to deal with the cover
art case: if the video has ended, use the audio PTS.
Also see commit f9a259e (and the commits referenced from there).
Trying to step over a segment boundary didn't work, and the video was
stuck at the end of the current chapter. At this point, both video and
audio of the segment has ended, and the segment switching code is going
to call seek() to go to the next segment (the part of the code in
run_playloop that uses end_is_chapter). However, this seek() is not
called if playback is paused, and the framestepping code always paused
before this code is run.
Move the framestepping code below the chapter switching code. The added
restart_playback condition makes sure the code is called only after at
least one video frame has been shown. Also don't reset the framestep
counter after seek. It's not needed, and removing it prevents full
unpausing when stepping over a segment boundary.
This also terminates playback when frame stepping at the end of the
file. The --keep-open option can be used to get the old behavior.
The OSX part of the Apple Remote was unmaintained for a long time and was not
working anymore. I tried to update the cookies to what the current versions of
OS X expect without much luck. I decided to remove it since Apple is not
including the IR receiver anymore in new hardware and it's clear that wifi
based remotes are the way to go.
A third party iOS app should be used in it's place. In the future we could look
into having a dedicated iOS Remote Control app like VLC and XBMC do.
The Linux side (`appleir.c`) was relatively tidy but it looks like LIRC can be
configured to work with any version of Apple Remote [1] and is more maintained.
[1] LIRC Apple Remote configs: http://lirc.sourceforge.net/remotes/apple/
Drawing the bar with vector drawings (instead with characters from the
OSD font) offers more flexibility and looks better. This also adds
chapter marks to the OSD bar, which are visible as small triangles on
the top and bottom inner border of the bar.
Change the default position of the OSD bar below the center of the
screen. This is less annoying than putting the bar directly into the
center of the view, where it obscures the video. The new position is
not quite on the bottom of the screen to avoid collisions with
subtitles.
The old centered position can be forced with ``--osd-bar-align-y=0``.
Also make it possible to change the OSD bar width/height with the new
--osd-bar-w and --osd-bar-h options.
It's possible that the new OSD bar renders much slower than the old
one. There are two reasons for this: 1. the character based bar
allowed libass to cache each character, while the vector drawing forces
it to redraw every time the bar position changes. 2., the bar position
is updated at a much higher granularity (the bar position is passed
along as float instead of as integer in the range 0-100, so the bar
will be updated on every single video frame).
gl_video.c contains all rendering code, gl_lcms.c the .icc loader and
creation of 3D LUT (and all LittleCMS specific code). vo_opengl.c is
reduced to interfacing between the various parts.
Makes the code a bit simpler to follow, at least in the "modern"
decoding path (update_video_nocorrect_pts() is used with old demuxers,
which don't return proper packets and need further parsing, so this code
looks less simple now).
There were complaints that ${fps} was printed as e.g. "23.98" instead of
"23.976". Since there's no way to format floats exactly _and_ in a user-
friendly way, just change the default precision for printing floats.
This way it's possible to retrieve correct information about video, like
actual width/height, which in general are available only after at least
one frame has been sent to the video output, such as dwidth/dheight.
mpv_identify.sh becomes a bit slower, because we let it decode enough
audio and video to fill the audio buffers and to send one frame to the
video output. Also, --playing-msg isn't shown anymore with --frames=0
(could be fixed by special-casing it, should this break any use cases).
Note that in some corner cases, like when the demuxer for some reason
returns lots of audio packets but no video packets at the start, but
video actually starts later, the --playing-msg will still be output
before video starts.
This has the same (useless) definition as frame stepping in audio-only
mode: one frame means one playloop iteration. (It's relatively useless,
because one playloop iteration has a random duration. But it makes
--frames=1 work, which is useful again.)
Add new properties "dwidth" and "dheight", which contain the video
size as known by the VO (not necessarily what the VO makes out of them,
i.e. without window scaling and panscan).
Some time ago, all old special-cased commands (like "volume 1" to change
volume by one) have been removed. These commands are still emulated
using simple text replacement. This emulation is done to not break
everyone's input.conf, especially because the input.conf provided by
standard mplayer* still uses the old commands.
Every use of a deprecated command prints a replacement warning, which
was visible only with -v. Make these warnings visible by default.
There's actually not much reason to do this, but since commands like
"volume 5 1" don't work anymore, it's better to be verbose about this.
Also simplify the replacement for "vo_fullscreen".
Normal text was set to gray foreground color. This didn't work for
terminals with white background.
Instead of setting a color for normal text, reset the color attributes.
This way, only errors and warnings are formatted differently.
Also change the default color for MSGL_HINT from bold white to yellow.
A recent change accidentally set the flags options to -1 (probably
confusing it with the defasult value?), which mistakenly set all flags
and rejected all option values (except 0).
We consider FFmpeg 1.x and Libav 0.9.x releases compatible. Support
for FFmpeg 0.9.x and Libav 0.8.x is considered infeasible and has been
dropped in the previous commits. The bits that break compatibility are
mainly the CodecID renaming (trivial, but would require nasty hacks
everywhere), the avcodec_encode_video2() function (missing in older
releases, mandatory in newer ones), and the resampler changes (older
releases miss lib{av,sw}resample, newer versions removed the
libavcodec resampler).
Remove some other compatibility bits that were needed to for releases
for which we drop support.
The comment about Libav 0.9 in compat/libav.h is incorrect and should
have been 0.8 (the symbol is present in Libav 0.9).
The old names have been deprecated a while ago, but were needed for
supporting older ffmpeg/libav versions. The deprecated identifiers
have been removed from recent Libav and FFmpeg git.
This change breaks compatibility with Libav 0.8.x and equivalent
FFmpeg releases.
Move them into per-instance structs. This should get rid of all global
variables in mplayer.c (not counting those referenced by cfg-mplayer.h).
In core/input/ar.c, just remove checking the slave_mode variable. I'm
not sure what this code was supposed to achieve, but slave mode is
broken, slave mode is actually infeasible on OSX (ar.c is completely OSX
specific), and the correct way of doing this would be to disable this
input device per command line switch.
Missing entries cause avcodec_descriptor_get() to return NULL, and in
turn mp_codec_from_av_codec_id() will return NULL. This shouldn't
happen, and avcodec_descriptor_get() returning NULL for a valid codec is
clearly a bug.
But make it more robust anyway, and use the decoder's name if this
happens, because I doubt maintainance of the AVCodecDescriptor table
in ffmpeg/Libav will always be perfect and reliable.
Latest nvidia drivers ignore the application setting, so this switch
makes even less sense than before. It's still possible to control this
with VO specific suboptions.
Separate the video output options from the big MPOpts structure and also only
pass the new mp_vo_opts structure to the vo backend.
Move video_driver_list into mp_vo_opts