Our code currently tries to link -lpthread and adds stuff like -D_REENTRANT
based on the target platform.
GCC actually supports to just pass a -pthread compiler and linker flag that
will automatically enable threading and define the correct symbols for the
platform, so let's try to just use that as our first choice.
clang also supports -pthread but it must be used only as a compiler flag,
so we also take care of that scenario with this commit.
One problem is that for example stdio functions won't restart syscalls
manually, and instead treat EINTR as an error. So passing SA_RESTART is
the only sane thing to do, unless you have special requirements, which
we don't.
VP9 packets can contain 2 frames in some video packets (from which 1
frame is invisible). Due to a design mismatch between libvpx and the
libavcodec vp9 decoder, libvpx can take the "full" packets, but lavc vp9
can not. The consequence is that we have to split the packets if we want
to feed them to the lavc codec.
This is not entirely correct yet: timestamp handling is missing.
--demuxer=lavf and ffmpeg native utilities have the same problem. We can
fix this only once the ffmpeg VP9 parser is fixed.
This used an unnamed union, which is allowed in GNU C and C11, but not
C99. This broke the build with some older compilers.
Replaces pull request #744.
This reverts commit 6e34b0ec1f.
There has always been an error message "proxy already has a listener" and
I couldn't reproduce where it is comming from until now. The display interface
already has a listener and we can't overwrite it. Now remove the code and avoid
this error message.
Conflicts:
video/out/wayland_common.c
This is the only function which actually used the time argument of
stream_check_interrupt(). Considering that the whole player freezes
anyway, this is not worth the complication.
Also generally reduce the maximum wait time due to timeout. Introduce
exponential backoff, which makes the first reconnect retries faster, but
still waits up to 500ms in the later retries.
For some reason, some files appear to have broken mp3 packets, or at
least in a form that libavcodec can't deal with. The audio in the sample
file in question could not be decoded using libavcodec.
The problematic file had variable packet sizes, and the libavcodec
decoder kept printing "mp3: Header missing" for each packet it was fed.
Remuxing with mkvmerge fixes the problem. The mp3 data is probably not
VBR, and remuxing resulted in fixed-size mp3 frames. So I don't know why
the sample file was muxed this way - it might just be incorrect.
The sample file had "libmkv 0.6.4" as MuxingApp (although I could not
get mkvinfo to print this element, maybe the file uses an incorrect
element ID), and "HandBrake 0.9.4" as WritingApp.
Note that the libmpg123 decoder does not have any issues with it. It's
probably more robust, because libmpg123 was made to decode whole mp3
files, not just single frames.
Fixes issue #742.
When using rotation with hw decoding, and the VO does not support
rotation, vf_rotate is attempted to be inserted. This will go wrong, and
after that it can't recover because a vf_scale filter was autoinserted.
Just removing all autoinserted filters before reconfig fixes this.
This avoids too many realloc() calls if the caller is appending to an
audo buffer. This case is actually quite noticeable when using something
that buffers a large amount of audio.
mpv was resizing to the same size before it went to fullscreen, we don't need to schedule a resize because the compositor will send a configure event with the new dimensions and thats when we should do it.
The stats were retrieved and written on every encode call, instead of
every encode call that actually returned a packet. ffmpeg.c also does it
this way, so it must be "more correct". Fixes 2-pass encoding.
Our own tables have size for only 8 chars, so these sequences must be
rejected. It seems strings of length 8 are still ok, because the code
uses memcmp and not strcmp, so still allow these.
Based on mplayer-svn commit r37129.
This might be a good idea in order to prevent queuing a frame too far in
the future (causing apparent freezing of the video display), or dropping
an infinite number of frames (also apparent as freezing).
I think at this point this is most of what we can do if the vdpau time
source is unreliable (like with Mesa). There are still inherent race
conditions which can't be fixed.
The strange thing about this code was the shift parameter of the
prev_vs2 function. The parameter is used to handle timestamps before the
last vsync, since the % operator handles negative values incorrectly.
Most callers set shift to 0, and _usually_ pass a timestamp after the
last vsync. One caller sets it to 16, and can pass a timestamp before
the last timestamp.
The mystery is why prev_vs2 doesn't just compensate for the % operator
semantics in the most simple way: if the result of the operator is
negative, add the divisor to it. Instead, it adds a huge value to it
(how huge is influenced by shift). If shift is 0, the result of the
function will not be aligned to vsyncs.
I have no idea why it was written in this way. Were there concerns about
certain numeric overflows that could happen in the calculations? But I
can't think of any (the difference between ts and vc->recent_vsync_time
is usually not that huge). Or is there something more clever about it,
which is important for the timing code? I can't think of anything
either.
So scrap it and simplify it.
vo_vdpau used a somewhat complicated and fragile mechanism to convert
the vdpau time to internal mpv time. This was fragile as in it couldn't
deal well with Mesa's (apparently) random timestamps, which can change
the base offset in multiple situations. It can happen when moving the
mpv window to a different screen, and somehow it also happens when
pausing the player.
It seems this mechanism to synchronize the vdpau time is not actually
needed. There are only 2 places where sync_vdptime() is used (i.e.
returning the current vdpau time interpolated by system time).
The first call is for determining the PTS used to queue a frame. This
also uses convert_to_vdptime(). It's easily replaced by querying the
time directly, and adding the wait time to it (rel_pts_ns in the patch).
The second call is pretty odd: it updates the vdpau time a second time
in the same function. From what I can see, this can matter only if
update_presentation_queue_status() is very slow. I'm not sure what to
make out of this, because the call merely queries the presentation
queue. Just assume it isn't slow, and that we don't have to update the
time.
Another potential issue with this is that we call VdpPresentationQueueGetTime()
every frame now, instead of every 5 seconds and interpolating the other
calls via system time. More over, this is per video frame (which can be
portantially dropped, and not per actually displayed frame. Assume this
doesn't matter.
This simplifies the code, and should make it more robust on Mesa. But
note that what Mesa does is obviously insane - this is one situation
where you really need a stable time source. There are still plenty of
race condition windows where things can go wrong, although this commit
should drastically reduce the possibility of this.
In my tests, everything worked well. But I have no access to a Mesa
system with vdpau, so it needs testing by others.
See github issues #520, #694, #695.
The --ass-styles option is implemented by calling ass_read_styles().
This function can take a codepage (so libass will use iconv to convert
it). This was implemented before our --subcp option was changed, and
this code was not updated. Now libass fails opening iconv, because
--subcp is not always (and not by default) a valid iconv codepage.
Just always pass NULL, which means the file passed to --ass-styles must
be in UTF-8. The --ass-styles option is a fringe option anyway (and will
destroy your subtitles), so having codepage support for it isn't
important at all.
demux_mf.c explicitly checks for the stream type to check whether images
are opened via pattern (mf://..., i.e. stream_mf.c) or directly. Of
course the stream type is not set to STREAMTYPE_MF if the stream is
wrapped through the cache, so it tried to open the pattern directly as
file, which failed.
Fix this by disabling caching for mf://. The cache doesn't make sense
here anyway, because each file is opened and closed every frame (perhaps
to avoid memory bloat).
It's possible that MPContext has a chapter list, but the demuxer
doesn't. In this case, accesing the chapter-metadata property would
lead to invalid accesses.
(This fixes the out of bound access, but in theory, the returned data
can still be incorrect, since MPContext chapters don't need to map
directly to demuxer chapters.)
This cd_info_t struct was practically unused. The only thing it did was
storing the track name of the form "Track %d" in a very roundabout way.
Remove it. (It made more sense when there was still CDDB support.)
Don't use an integer division to get the time, since that would round on
second boundaries. Also round up the time by sector size. Seeking rounds
down due to alignment constraints, but if we round up the time, we can
make it land on the exact destination sector.
This fixes that the track change code printed the previous track when
seeking by chapter.
Report the time for each chapter (tracks are treated as chapters). This
allows us to get rid of the "old" chapter mechanism, and also behaves
better with the frontend.
This makes assumptions about the audio formats, but that format is
hardcoded anyway in the rawaudio demuxer defaults (and always was).
This was usually handled at the end of the switch statement, so if
something returns from the function before that, the event has to be
freed explicitly.
