The plan is to make the whole OSD thread-safe, and we start with this.
We just put locks on all entry points (fortunately, dec_sub.c and all
sd_*.c decoders are very closed off, and only the entry points in
dec_sub.h let you access it). I think this is pretty ugly, but at least
it's very simple.
There's a special case with sub_get_bitmaps(): this function returns
pointers to decoder data (specifically, libass images). There's no way
to synchronize this internally, so expose sub_lock/sub_unlock functions.
To make things simpler, and especially because the lock is sort-of
exposed to the outside world, make the locks recursive. Although the
only case where this is actually needed (although trivial) is
sub_set_extradata().
One corner case are ASS subtitles: for some reason, we keep a single
ASS_Renderer instance for subtitles around (probably to avoid rescanning
fonts with ordered chapters), and this ASS_Renderer instance is not
synchronized. Also, demux_libass.c loads ASS_Track objects, which are
directly passed to sd_ass.c. These things are not synchronized (and
would be hard to synchronize), and basically we're out of luck. But I
think for now, accesses happen reasonably serialized, so there is no
actual problem yet, even if we start to access OSD from other threads.
This readds a more or less completely new dvdnav implementation, though
it's based on the code from before commit 41fbcee. Note that this is
rather basic, and might be broken or not quite usable in many cases.
Most importantly, navigation highlights are not correctly implemented.
This would require changes in the FFmpeg dvdsub decoder (to apply a
different internal CLUT), so supporting it is not really possible right
now. And in fact, I don't think I ever want to support it, because it's
a very small gain for a lot of work. Instead, mpv will display fake
highlights, which are an approximate bounding box around the real
highlights.
Some things like mouse input or switching audio/subtitles stream using
the dvdnav VM are not supported.
Might be quite fragile on transitions: if dvdnav initiates a transition,
and doesn't give us enough mpeg data to initialize video playback, the
player will just quit.
This is added only because some users seem to want it. I don't intend to
make mpv a good DVD player, so the very basic minimum will have to do.
How about you just convert your DVD to proper video files?
Use sh_stream over sh_sub. Use dec_sub (and mpctx->d_sub) instead of the
stream header. This aligns the subtitle code with the recent audio and
video refactoring.
sh_sub still has the decoder context, though. This is because we want to
avoid reinit when switching segments with ordered chapters. (Reinit is
fast, except for creating the ASS_Renderer, which in turn triggers
fontconfig.) Not sure how much this matters, though, because the initial
segment switch will lazily initialize the decoder anyway.
This means the direct libass usage can be removed from command.c, and no
weird hacks for retrieving the ASS_Track are needed.
Also fix a bug when using this feature with ordered chapters.
If a subtitle is external, read it completely and add all subtitle
events in advance when the subtitle track is selected. This is done
for text subtitles only. (Note that subreader.c and subtitles loaded
with libass are different and don't have anything to do with this
commit.)
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.
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.
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.
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.
Remove VFCTRL_DRAW_OSD, VFCAP_EOSD_FILTER, VFCAP_EOSD_RGBA, VFCAP_EOSD,
VOCTRL_DRAW_EOSD, VOCTRL_GET_EOSD_RES, VOCTRL_QUERY_EOSD_FORMAT.
Remove draw_osd_with_eosd(), which rendered the OSD by calling
VOCTRL_DRAW_EOSD. Change VOs to call osd_draw() directly, which takes
a callback as argument. (This basically works like the old OSD API,
except multiple OSD bitmap formats are supported and caching is
possible.)
Remove all mentions of "eosd". It's simply "osd" now.
Make OSD size per-OSD-object, as they can be different when using
vf_sub. Include display_par/video_par in resolution change detection.
Fix the issue with margin borders in vo_corevideo.
This fixes that vo_xv didn't display text subtitles correctly when
using anamorphic video. It didn't pass the aspect information to the
subtitle renderer. Also, try to render OSD correctly with respect to
aspect ratio settings: on vo_xv, the OSD is rendered into the video,
and needs to be "stretched" too when playing anamorphic video. When
the -monitorpixelaspect option is used, even with VOs such as vo_opengl
the OSD has to be rendered with that aspect ratio.
As preparation for future commits, replace the weird vsfilter_scale
value with a somewhat more sensible video_par member.
Also, struct mp_eosd_res is a better place for the aspect ratio
parameters, as OSD needs this too.
Use osd_draw_on_image() directly in vo_lavc, which fixes aspect ratio
issues as well.
Instead, sd_lavc.c and spudec.c (the two image sub decoders) always
output indexed/paletted images. For this purpose, add SUBBITMAP_INDEXED,
and convert the subs to RGBA in img_convert.c instead. If a VO is used
that supports the old OSD format only, the indexed bitmaps are converted
to the old OSD format by abusing spudec.c in a similar way sd_lavc.c
used to do.
The main reason why spudec.c is used is because the images must not only
be converted to the old format, but also properly scaled, cropped, and
aligned (the asm code in libvo/osd.c requires this alignment).
Remove support for the old format (packed variant) from the OpenGL VOs.
(The packed formats were how the actual OSD format was handled in some
GPU-driven VOs for a while.)
Remove all conversions from old to new formats. Now all subtitle
decoders and OSD renderers produce the new formats only.
Add an evil hack to convert the new format (scaled+indexed bitmaps) to
the old format. It creates a new spudec instance to convert images to
grayscale and to scale them. This is temporary for VOs which don't
support new OSD formats yet (vo_xv, vo_x11, vo_lavc).
Move sub-bitmap definitions from dec_sub.h to sub.h. While it's a bit
odd that OSD data structures are in a file named sub.h, it's definitely
way too strange to have them in a file about subtitle decoding. (Maybe
sub.h/.c and the sub/ directory should be split out and renamed "osd"
at a later point.)
Remove including ass_mp.h (and the libass headers) where possible.
Remove typedefs for mp_eosd_res and sub_bitmaps structs.
Store a mp_eosd_res struct in osd_state instead of just w/h. Note that
sbtitles might be rendered using different sizes/margins when filters
are involved (the subtitle renderer is not supposed to use the OSD res
directly, and the "dim" member removed in the previous commit is
something different).
Passing parameters from caller to subtitle renderer was done by
temporarily setting certain members in the osd_state struct (which for
all practical purposes are as good as global variables). This was the
only purpose of these members.
Rather than using such a messy way to pass parameter, put these into a
struct sub_render_params. The struct was already introduced in earlier
commits, and this commit just removes the parameter passing hack.
Before this commit, vf_vo.c and vf_ass.c were manually calling the
subtitle decoder to retrieve images to render. In particular, this
circumvented the sub-bitmap conversion & caching layer in sub.c.
Change this so that subtitle decoding isn't special anymore, and draws
all subtitles with the normal OSD drawing API.
This is also a step towards removing the need for vf_ass auto-insertion.
In fact, if auto-insertion would be disabled now, VOs with "old" OSD
rendering could still render ASS subtitles in monochrome, because
there is still ASS -> old-OSD bitmap conversion in the sub.c mechanism.
The code is written with the assumption that the subtitle rendering
filter (vf_ass) can render all subtitle formats. Since vf_ass knows the
ASS format only, rendering image subs (i.e. RGBA subs) with it simply
fails. This means that with vo_xv (vf_ass auto-inserted), image subs
wouldn't be rendered. Use a dumb hack to disable rendering subs with a
filter, if we detect that the subs are not in ASS format. (Trying to
render the subs first would probably result in purging the conversion
cache on every frame.)
Before this commit, the OSD was drawn using libass, but the resulting
bitmaps were converted to the internal mplayer OSD format. We want to
get rid of the old OSD format, because it's monochrome, and can't even
be rendered directly using modern video output methods (like with
OpenGL/Direct3D/VDPAU).
Change it so that VOs can get the ASS images directly, without
additional conversions. (This also has the consequence that the OSD can
render colors now.) Currently, this is vo_gl3 only. The other VOs still
use the old method. Also, the old OSD format is still used for all VOs
with DVD subtitles (spudec).
Rewrite sub.c. Remove all the awkward flags and bounding boxes and
change detection things. It turns out that much of that isn't needed.
Move code related to converting subtitle images to img_convert.c. (It
has to be noted that all of these conversions were already done before
in some places, and that the new code actually makes less use of them.)
One sub_bitmaps struct could contain either a libass ASS_Image list, or
a mplayer native list of sub-bitmaps. This caused code duplication in
vo_vdpau.c and bitmap_packer.c.
Avoid this by creating such a sub_bitmap array even with libass. This
basically copies the list and recreates it in mplayer's native format.
It gets rid of the code duplication, and will make implementing extended
subtitle and OSD rendering in other VOs easier.
Also do some cosmetic changes and other preparations for the following
commits.
Support passing bitmap subtitles to VOs in full RGBA color, and
implement this for libavcodec-decoded subtitle formats on decoding
side and vo_vdpau on display side. Currently this is enabled for PGS
(blu-ray) and DVB subtitles.
VDPAU seems to have sampling issues similar to known GL ones when
drawing a sub-rectangle from a larger texture with scaling, where
adjacent pixels outside the specified source rectangle affect the
result. As the bitmap subtitles may be scaled, add padding support to
the bitmap packer code.
In principle, this could be used for colored DVD subtitles too.
However, the libavcodec DVD decoder lacks parts of the resolution and
palette handling that are present in spudec.c.
Conflicts:
libvo/vo_gl.c
sub/dec_sub.h
sub/sd_lavc.c
This was destroyed by Uoti Urpala in commit "subs: always use sub...".
Features should be either kept working or completely removed, but not
just crippled, which only inflates the code and frustrates users.
vo_vdpau and vo_gl cache the last subtitle bitmaps uploaded to video
card in case they stay the same over multiple frames. Detecting
whether the bitmaps have changed and should be re-uploaded was
somewhat fragile. Change the VO API to provide a bitmap ID which can
be compared with what the VO has to determine whether a new upload of
the bitmaps is needed.
Conflicts:
libvo/vo_gl.c
Note: the changes for vo_gl.c were not merged. Instead, eosd_packer is
modified to use the new way of detecting EOSD changes. This takes care
of vo_gl, vo_gl3 and vo_direct3d, which all render EOSD. They don't
need to be updated in turn.
Remove subtitle selection code setting osd->ass_track directly and
vf_ass/vf_vo code rendering the track directly with libass. Instead,
do track selection and rendering with dec_sub.c functions.
Before, mpctx->set_of_ass_tracks[] contained bare libass tracks
generated from external subtitle files. For use with dec_sub.c, it now
contains struct sh_sub instances with decoder already initialized.
This commit breaks the sub_step command ('g' and 'y' keys) for
libass-rendered subtitles. It could be fixed, but it's so useless -
especially as with the existing implementation there's no practical
way to get subtitle delay back to normal after using it - that I
didn't bother.
Conflicts:
command.c
mp_core.h
mplayer.c
To draw libass subtitles, the code used ASS_Renderer objects created
in vf_vo (VO rendering) or vf_ass. They were destroyed and recreated
together with the video filter chain. Change the code to use a single
persistent renderer instance stored in the main osd_state struct.
Because libass seems to misbehave if fonts are changed while a
renderer exists (even if ass_set_fonts() is called on the renderer
afterwards), the renderer is recreated after adding embedded fonts.
The known benefits are simpler code and avoiding delays when switching
between timeline parts from different files (libass fontconfig
initialization, needed when creating a new renderer, can take a long
time in some cases; switching between files rebuilds the video filter
chain, and this required recreating the renderers). On the other hand,
I'm not sure whether this could cause inefficient bitmap caching in
libass; explicitly resetting the renderer in some cases could be
beneficial. The new code does not keep the distinction of separate
renderers for vsfilter munged aspect vs normal; this means that
changing subtitle tracks can lose cache for the previous track.
The new code always sets some libass parameters on each rendering
call, which were previously only set if they had potentially changed.
This should be harmless as libass itself has checks to see if the
values differ from previous ones.
Conflicts:
command.c
libmpcodecs/vf_ass.c
libmpcodecs/vf_vo.c
mplayer.c
sub/ass_mp.c
Add a framework for subtitle decoder modules that work more like
audio/video decoders do, and change libass rendering of demuxed
subtitles to use the new framework.
The old subtitle code is messy, with details specific to handling
particular subtitle types spread over high-level code. This should
make it easier to clean things up and fix some bugs/limitations.
wm4
Uoti Urpala
Uoti Urpala