This change affects vf_lavfi. Until recently, libavfilter was not
colorspace aware at all. This changed with the addition of colorspace
fields to AVFrame. libavfilter's vf_scale picks them up (as of recent
ffmpeg git). Since this support is still kind of wonky and not part of
the normal format negotiation, this won't set the correct output
colorspace, though.
Not adding a separate test for HAVE_AVFRAME_COLORSPACE. This is slightly
unclean, but on the other hand adding an explicit test seems like a
waste of effort.
The symptom was that "-vf scale,format=rgba" broke the vsfilter
colorspace hack in sd_ass. vf->reconfig is allowed to overwrite its
input parameter for convenience (maybe that was a bad idea).
Windows doesn't send WM_MOUSELEAVE by default unless you ask it to;
request tracking for leave events when the mouse enters the window (or is
moved).
Tracking is automatically de-activated once the mouse leaves the window,
so we have to re-request it every time the mouse re-enters the window.
It appears the API requires you to cover all plane data with AVBuffers
(that is, one AVBuffer per plane in the most general case), because
certain code can make certain assumptions about this. (Insert rant
about how this is barely useful and increases complexity and potential
bugs.) I don't know any cases where the current code actually fails,
but we want to follow the API, so do it anyway.
Note that we don't really know whether or not planes are from a single
memory allocation, so we have to assume the most general case and create
an AVBuffer for each plane. We simply assume that the data is padded to
the full stride in the last image line. All these extra dummy references
are stupid, but the code might become much simpler once we only support
libavcodec versions with refcounting and can use AVFrame directly.
In general, this warning can hint to actual bugs. We don't enable it
yet, because it would conflict with some unmerged code, and we should
check with clang too (this commit was done by testing with gcc).
Doing "mpv --vo=opengl:lscale=help" now lists possible scalers and
exits. The "backend" suboption behaves similar. Make the "stereo"
suboption a choice, instead of using magic integer values.
Commit 6ab2eeb attempted to fix it on Cygwin, but now it broke on MinGW
in turn. Don't think too hard about it and just remove the code. (vo.c
already prints the video rectangle anyway.)
Nothing should change from user perspective.
mpv --vo=opengl:help now works.
Remove the vo_opengl inline help text. The new code can list option
names for you, but that's it. Refer to the manpage if you have trouble.
For some reason, both m_config and m_struct are somewhat similar, except
that m_config is much more powerful. m_config is used for VOs and some
other things, so to unify them. We plan to kick out m_struct and use
m_config for everything. (Unfortunately, m_config is also a bit more
bloated, so this commit isn't all that great, but it will allow to
reduce the option parser mess somewhat.)
This commit also switches all video filters to use the option macros.
One reason is that m_struct and m_config, even though they both use
m_option, store the offsets of the option fields differently (sigh...),
meaning the options defined for either are incompatible. It's easier to
switch everything in one go.
This commit will allow using the -vf option parser for other things,
like VOs and AOs.
When the cursor was in the window border, it could be hidden but it
wouldn't appear again, since mpv doesn't process mouse input there.
The code used ShowCursor, which is a horrid stateful API designed for
mouseless Win16 systems that incremented or decremented a global counter
to keep track of how many applications needed to display a special
cursor (like a busy cursor.) Replace that with a simple flag, handle
WM_SETCURSOR and use SetCursor(NULL) to hide the mouse cursor, but only
when the mouse is in the client area. DefWindowProc will set the correct
cursor on the border as long as it isn't hidden with ShowCursor.
PowerPoint also uses SetCursor(NULL) to hide the cursor when showing a
presentation, so it's probably safe.
See http://blogs.msdn.com/b/oldnewthing/archive/2009/12/17/9937972.aspx
This change which also flipse the coordinate system of the view, greatly
simplifies the mouse event handling code.
There are still some uglities mostly related to the cursor visibility
code. For instance the core doesn't show the cursor when it receives a mouse
leave event.
This was more roundabout than expected, since it looks like the framework
caches isMovabileByWindowBackground so in mpv's case it's needed to set it
with setMovableByWindowBackground.
This commit removes the pointer to the single different structures for input
and window and puts them as anonymous structures inside the wayland_state
structure.
This has the disadvantage of passing the substructure to the listeners, but the
advantage is that we don't have to allocate them and check for NULL pointers.
This makes it more reliable and easier to follow.
The vo_wayland_fullscreen handles resizing for the video, because the video
could still be in fullscreen mode and resizing it in gl_wayland could make it
grow or shrink.
There was a MPOpts fullscreen field, a mp_vo_opts.fs field, and
VOFLAG_FULLSCREEN. Remove all these and introduce a
mp_vo_opts.fullscreen flag instead.
When VOs receive VOCTRL_FULLSCREEN, they are supposed to set the
current fullscreen mode to the state in mp_vo_opts.fullscreen. They
also should do this implicitly on config().
VOs which are capable of doing so can update the mp_vo_opts.fullscreen
if the actual fullscreen mode changes (e.g. if the user uses the
window manager controls). If fullscreen mode switching fails, they
can also set mp_vo_opts.fullscreen to the actual state.
Note that the X11 backend does almost none of this, and it has a
private fs flag to store the fullscreen flag, instead of getting it
from the WM. (Possibly because it has to deal with broken WMs.)
The fullscreen option has to be checked on config() to deal with
the -fs option, especially with something like:
mpv --fs file1.mkv --{ --no-fs file2.mkv --}
(It should start in fullscreen mode, but go to windowed mode when
playing file2.mkv.)
Wayland changes by: Alexander Preisinger <alexander.preisinger@gmail.com>
Cocoa changes by: Stefano Pigozzi <stefano.pigozzi@gmail.com>
This was bad, because it was the only aspdat member updated by
vo_get_src_dst_rects() instead of vo_reconfig(). Now it isn't
accessed anymore, so remove it.
Until now, only formats directly supported by OpenGL were supported.
This excludes various permutations of 8-bit RGB[A|0]. But we can simply
permutate the color channels in the shader, so do that. This also adds
support for all these weird RGB0 formats.
Note that we could use libavutil's pixfmt list instead of the
mp_packed_formats array, but trying to decrypt the pixfmt info would
probably end in pain, so this array with duplicated information is
actually better and shorter.
Note: I didn't actually test whether the alpha components are reproduced
correctly with alpha formats.
Image parameters like colorspace, color levels, and chroma location are
generally less important, and many filters don't set them correctly.
Force them instead in the generic VF code, which is probably better and
more convenient over all. So we designate this is a proper solution,
instead of a dirty hack.
This splits the monolithic mp_image_swscale() function into a bunch of
functions and a context struct. This means it's possible to set
arbitrary parameters (e.g. even obscure ones without getting in the
way), and you don't have to create the context on every call.
This code is preparation for removing duplicated libswscale API usage
from other parts of the code.
libswscale doesn't seem to require this (anymore?), and libavfiltert's
vf_scale doesn't do it either. Moreover, this wasn't done for most other
subsampled formats, not even very old ones. So just remove it.
(It'd be quite easy to align on chroma boundaries with all pixel
formats, though.)
Until now, vf_scale only tried formats listed in the outfmt_list array.
Extend this and try every pixel format supported by mpv if trying
outfmt_list doesn't lead to success.
Also add some checks whether swscale really supports a given input or
output format. This was implicitly done with outfmt_list before.
ffmpeg's and the internal palette format used to have different
endianess, but that is not the case anymore. This code was forgotten
when that change was made.
I guess this code was supposed to handle cases like drawing RGBA as ARGB
by offsetting it by 1 byte.
The code didn't make any sense, though. It used to make sense before mpv
switched internal pixel formats from FourCCs to a simple enum. With the
FourCCs, "fmt | 128" selected the big endian version of a format. Of
course this doesn't work this way with the new pixel formats. It just so
happens that there are no formats with whose values match
IMGFMT_RGB32|128 or IMGFMT_BGR32|128, so this code was inactive.
All involved pixel formats seem to play fine on my setup (though it's
little endian only), and the code strictly matches the mpv pixel formats
against the format of the X image, so I'm not quite sure why this code
was there in the first place.
The original commit that added this was b333ae1 (svn 21602):
Support for different endianness on client and server with -vo x11
Instead of handling colorspaces with VFCTRLs/VOCTRLs, make them part of
the normal video format negotiation. The colorspace is passed down like
other video params with config/reconfig calls.
Forcing colorspaces (via the --colormatrix options and properties) is
handled differently too: if it's changed, completely reinit the video
chain. This is slower and requires a precise seek to the same position
to perform an update, but it's simpler and less bug-prone. Considering
switching the colorspace at runtime by user-interaction is a rather
obscure feature, this is a good change.
The colorspace VFCTRLs and VOCTRLs are still kept. The VOs rely on it,
and would have to be changed to get rid of them. We'll do that later,
and convert them incrementally instead of in one go.
Note that controlling the output range now always works on VO level.
Basically, this means you can't get vf_scale to output full-range YUV
for whatever reason. If that is really wanted, it should be a vf_scale
option. the previous behavior didn't make too much sense anyway.
This commit fixes a few bugs (such as playing RGB video and converting
that to YUV with vf_scale - a recent commit broke this and forced the
VO to display YUV as RGB if possible), and might introduce some new
ones.
Calculate the aspect ratio in vo_config, when we get the window size and in the
inside the resize function we calculate the aspect ratio of the output in order
to determine if we have to change the height or the width of the video.
If the ratio of the output is bigger than the ratio of the video then we have
to set the width accordingly and if the ratio is smaller we change the size.
But only if no resize edges are passed, because this indicates that we want to
change the windows state instead of just a simple resize and the video should
not grow bigger than the requested size.
aspdat.asp is a problem, because it's updated when the VO calls
vo_get_src_dst_rects(). Nothing guarantees that the value has been
updated when the w32 code accesses it.
Instead, use the aspect vo_w32_config() was called with.
Guess the colorspace directly in mpcodecs_reconfig_vo(), instead of in
set_video_colorspace(). The difference is that the latter function just
makes the video filter chain (and VOs) force the detected colorspace,
and then throws it away, while the former is a bit more general and
central. Not really a big difference and it doesn't matter much in
practice, but it guarantees that there is no internal disagreement about
the colorspace.
This removes the dependency on DEMUXER_TYPE_* and the file_format
parameter from the stream open functions.
Remove some of the playlist handling code. It looks like this was
needed only for loading linked mov files with demux_mov (which was
removed long ago).
Delete a minor bit of dead network-related code from stream.c as well.
Delete demux_avi, demux_asf, demux_mpg, demux_ts. libavformat does
better than them (except in rare corner cases), and the demuxers have
a bad influence on the rest of the code. Often they don't output
proper packets, and require additional audio and video parsing. Most
work only in --no-correct-pts mode.
Remove them to facilitate further cleanups.
This fixes compilation with broken libcs, like on Cygwin. C99
absolutely requires long double and associated functions like expl,
even if long double is double. But newlib (used by cygwin) omits
declaration for these if long double is equivalent to double.
The extra precision is not needed here, so remove it to make life
easier for the single person using mpv with cygwin.
For some reason mp_fifo specifically handled double clicks, and other
than that was a pointless wrapper around input.c functionality.
Move the double click handling into input.c, and get rid of mp_fifo. Add
some compatibility wrappers, because so much VO code uses these
functions. Where struct mp_fifo is still used it's just a casted
struct input_ctx.
Making key up events implicit was sort-of a nice idea, but it's too
tricky and unreliable and makes the key lookup code (interpret_keys())
hard to reason about. See e.g. previous commit for subtle bugs and
issues this caused.
Make key-up events explicit instead. Add key up events to all VOs.
Any time MP_KEY_STATE_DOWN is used, the matching key up event must
use MP_KEY_STATE_UP.
Rewrite the key lookup code. It should be simpler and more robust now.
(Even though the LOC increases, because the new code is less "compact".)
Handling SC_MONITORPOWER doesn't seem to prevent the screen from
dimming. The recommended way to do this in Windows XP and Vista is to
call SetThreadExecutionState with ES_DISPLAY_REQUIRED. Windows 7 also
has the PowerCreateRequest/PowerSetRequest/PowerClearRequest APIs but
they're probably too complicated for this task.
Also, implement mouse leave events for X11. But evne on other
platforms, these events will be generated if mouse crosses a section's
mouse area boundaries within the mpv window.
Before this commit, mouse movement events emitted a special command
("set_mouse_pos"), which was specially handled in command.c. This was
once special-cased to the dvdnav and menu code, and did nothing after
libmenu and dvdnav were removed.
Change it so that mouse movement triggers a pseudo-key ("MOUSE_MOVE"),
which then can be bound to an arbitrary command. The mouse position is
now managed in input.c. A command which actually needs the mouse
position can use either mp_input_get_mouse_pos() or mp_get_osd_mouse_pos()
to query it. The former returns raw window-space coordinates, while the
latter returns coordinates transformed to OSD- space. (Both are the same
for most VOs, except vo_xv and vo_x11, which can't render OSD in
window-space. These require extra code for mapping mouse position.)
As of this commit, there is still nothing that uses mouse movement, so
MOUSE_MOVE is mapped to "ignore" to silence warnings when moving the
mouse (much like MOUSE_BTN0).
Extend the concept of input sections. Allow multiple sections to be
active at once, and organize them as stack. Bindings from the top of
the stack are preferred to lower ones.
Each section has a mouse input section associated, inside which mouse
events are associated with the bindings. If the mouse pointer is
outside of a section's mouse area, mouse events will be dispatched to
an input section lower on the stack of active sections. This is intended
for scripting, which is to be added later. Two scripts could occupy
different areas of the screen without conflicting with each other. (If
it turns out that this mechanism is useless, we'll just remove it
again.)
Some functions (avcol_spc_to_mp_csp() etc.) used libavcodec enum types
as parameters. Remove these in order to get rid of the avcodec.h
include statement. This prevents that avcodec.h is recursively
included by dozens of files. Fix mp_image.c, which used the header
without explicitly including avcodec.h.
Use the video decoder chroma location flags and render chroma locations
other than centered. Until now, we've always used the intuitive and
obvious centered chroma location, but H.264 uses something else.
FFmpeg provides a small overview in libavcodec/avcodec.h:
-----------
/**
* X X 3 4 X X are luma samples,
* 1 2 1-6 are possible chroma positions
* X X 5 6 X 0 is undefined/unknown position
*/
enum AVChromaLocation{
AVCHROMA_LOC_UNSPECIFIED = 0,
AVCHROMA_LOC_LEFT = 1, ///< mpeg2/4, h264 default
AVCHROMA_LOC_CENTER = 2, ///< mpeg1, jpeg, h263
AVCHROMA_LOC_TOPLEFT = 3, ///< DV
AVCHROMA_LOC_TOP = 4,
AVCHROMA_LOC_BOTTOMLEFT = 5,
AVCHROMA_LOC_BOTTOM = 6,
AVCHROMA_LOC_NB , ///< Not part of ABI
};
-----------
The visual difference is literally minimal, but since videophiles
apparently consider this detail as quality mark of a video renderer,
support it anyway. We don't bother with chroma locations other than
centered and left, though.
Not sure about correctness, but it's probably ok.
The filter chain and the video ouputs have config() functions. They are
strictly limited to transfering the video size and format. Other
parameters (like color levels) have to be transferred separately.
Improve upon this by introducing a separate set of reconfig() functions,
which use mp_image_params to carry format parameters. This struct
contains all image format related parameters from config(), plus
additional parameters such as colorspace.
Change vf_rotate to use it, as well as vo_opengl. vf_rotate is just
an example/test case, but vo_opengl will need it later.
The intention is also to get rid of VOCTRL_SET_YUV_COLORSPACE. This
information is now handed to the VOs via reconfig(). The getter,
VOCTRL_GET_YUV_COLORSPACE, will still be needed though.
Normally, we assume that IMGFMT_PAL8 always has a palette allocated
in plane 1. But there may be corner cases in ffmpeg where it doesn't
(namely pseudo-pal stuff).
A redraw forces recalculation of panscan and other stuff not accounted for in
the resize_redraw codepath. This is actually a hack but works really well in
my tests.
Thanks @wm4 and @Cpuroast for the idea.
Fixes#86
[ci skip]
Seems like a completely unnecessary complication. Instead, always add a
1 byte padding (could be extended if a caller needs it), and clear it.
Also add some documentation. There was some, but it was outdated and
incomplete.
Recent work in the OS X parts of the code started using clang's support for
Obj-C's support for Literals and Subscripting. These particular language
features remove a lot of boilerplate code and allow to interact with
collections as consicely as one would do in scripting languages like Ruby or
Python.
Even if these are compiler features, Subscripting needs some runtime support.
This is provided with libarclite (coming with the compiler), but we need to
add the proper method definitions since the 10.7 SDK headers do not include
them. That is because 10.7 shipped before this language features.
This will cause some warnings when compiling with the 10.7 SDK because the
commit also redefines BOOL to make autoboxing/unboxing of BOOL literals to
work.
If you need to test this for whatever reason on 10.8, just pass in the correct
SDK to configure's extra cflags:
./configure --extra-cflags='-mmacosx-version-min=10.7 -isysroot /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.7.sdk'
Fixes#117
Doing this makes the encoder force the same pict type as original, which
is often not even possible. Rather let the codec decide!
As there is no documented value to mean "decoder shall pick", I rather
save/restore the default value filled by libavcodec.
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.
This make the intention more apparent, and some VOs are actually using
true instead of VO_TRUE in some places. Hopefully this changes makes it
less confusing (instead of more).
The C99 constants true/false are defined to 1/0 as well, so this commit
doesn't actually change anything.
This code calculates the source/display video rectangle for scaling with
most VOs. It's responsible for clipping the display rectangle against
the screen and adjusting the source rectangle accordingly.
Until now, it assumed that the video was centered on the screen. Change
this so that any rectangle is possible. Basically, the clipping is
extended to two sides (e.g. left and right), instead of handling both at
the same time.
The rounding behavior slightly changes. It seems to be slightly better
than before. On the other hand, the video is not strictly centered
anymore (due to different rounding on either side). When using panscan
controls, the video can "jitter" by 1 or 2 pixels around the center as
the panscan value is changed.
When the displayed image is cropped in Y direction (like using panscan
controls when playing 4:3 video on a 16:9 monitor), and separated
scaling is used, the texture size for the FBO holding the intermediate
result was calculated incorrectly. This could lead to artifacts, which
were quite apparent with extreme scale factors.
Actually, the size of that texture is OK, but the texture shouldn't be
used to hold the complete scaled image. Instead, it should be used for
the visible part of the image only. Because separate scaling works by
scaling in Y direction first, it's still fine to scale the image on the
full image width on the first pass. This helps avoiding artifacts on
the left/right border of the image when scaling in X direction, as the
scaler will try to fetch pixels from beyond the border. (The left border
is still kind of fine, but the right border will fetch garbage, unless
the texture is strictly sized, or explicit clamping is added to the
shader. Too much trouble, so using the full image width is simpler.)
Also fix some issues with no-npot mode, which enables use of power-of-2
textures. Maybe this mode isn't really useful anymore (modern hardware
is faster with smaller non-power-of-2 textures), but keep it for now.
The use of filters prior to PNG compression can greatly improve
compression ratio, with "mixed" (ImageMagick calls it "adaptive")
typically achieving the best results.
This is quite similar to the previous commit.
Untested. I'm not sure if this is how it's supposed to work. At least
--no-stop-screensaver should work in any case.
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.
`enterFullScreenMode:withOptions:` creates another window so set the level on
both the windowed window and current window.
Also remove NSFullScreenModeWindowLevel as it seems to be superfluous.
Now it properly hits the "0 times displayed" case when frames get
skipped; this means the candidate frame for the case the next frame is
"long" is set properly.
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.)
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.
Now that Cocoa's input handling is done on a separate thread from the playloop
it is ridicolously simple to have longer asynchronous sleeps when paused.
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.
Originally, the header wasn't supposed to contain random compatibility
stuff, but now all that is printed with -v. Add a hack to skip it and
to reduce the noise.
This could lead to quite visible artifacts when using an appropriate ICC
and float FBOs. The float FBOs allow storing out of range values, and my
guess is that the rest of the precessing chain elevated these out of
range values, resulting in artifacts.
Autohide the menubar and/or dock only if they are present in the screen the
player is going to go fullscreen into. I thought the GUI would handle this for
me when I switched 0057aa476 but lack of hardware to test made me embarass
myself yet again.
I reimplemented this feature with nicer code and behaviour. The code checks
separately wether to hide menubar and dock separatly, while the old code used
a single check possibly hiding stuff without need.
Added the key checks as a some category additions to NSScreen for readability.
This takes an approach similar to the wayland OpenGL backend. VOFLAG_HIDDEN
flag semantics doesn't mean "hide the window" but is simply ever used only to
do detection of available OpenGL extensions. On OSX it's possibile to
accomplish this task just by creating the OpenGL context without attaching
it to a drawable.
Using `enterFullScreenMode:withOptions:` with a screen handle than the current
screen doesn't hide the current window in the current screen. This is a bug in
Cocoa (preparing an isolated test case and sending the rdar later).
To work around this, manually hide/show the window that the toolkit should
hide/show for us.
This bug was the result of crappy position detection in the previous code
combined with the commits moving autohide delay out of the cocoa backend and
into the core.
The hit detection was improved and now takes also account of interactions with
the Dock and Menubar. Moreover VOCTRL_SET_CURSOR_VISIBILITY now has an effect
only if the mouse position matches with this improved hit detection. This means
that both interaction with the Dock and Menubar are considered as well as
moving the mouse inside another screen.
This removes a bit of ugly code and bookeeping which is never bad. `drawRect`
needs to guard against different window instances since in fullscreen the view
is wrapped in a fullscreen window provided by the toolkit (a instance of
NSFullScreenWindow to be precise).
The event handling was moved to the view so that it can still get all the
events when in the fullscreen window. Ideally these should be moved to
some NSResponder subclass within macosx_application and made available even
when no window is present. I refrained from this because "small steps".
At this point 10.6 is pretty old and we don't want to supporting old platforms.
I'm killing all the 10.6 compatibility code before doing more refactorings.
Next commits will also use newer Objective-C syntax such as literals and
@autoreleasepool.
0 is invalid. The intention of the code turning off any additional
alignment, so we need 1.
Change a comment: obviously we don't try to set alignment parameters
etc.to handle stride correctly, and instead do everything by row.
This probes and prints the depth of some texture formats with the help
of a FBO. By default it tests the format used for scaling, as well as
the format used for dithering and the 3D LUT (if any of these are
enabled).
The output is visible only with -v. Some representative values are
probed, and the difference of input and output value is printed as hex-
float. Hex-floats are used because they make the implied precision more
obvious. Originally I wanted to do some more sophisticated guessing of
the implied depth/precision for more user-friendly reporting, but then
I decided that printing raw data is better for debugging, especially if
things go wrong.
This does not try to disable any functionality and does not print any
warnings if the depth is lower than what it should be.
This might be better with dumb shader compilers, which won't vectorize
this to a single vector-division, assuming the hardware does have such
an instruction. Affects "bicubic_fast" scale mode only.
The internal texture format GL_RED is typically 8 bit, which is clearly
not good enough for the new dither matrix. The idea was to use a float
texture format, but this was somehow "forgotten". Use GL_R16, since
16 bit textures are more robust, and provide more precision for the
same memory usage.
Change how the offset for centering the dither matrix is applied. This
is needed for making it possible to round up values to the target depth.
Before this commit, this changed the output even if the input was exact
and input and output depth were the same, which is not really what you
want. Now it doesn't do that anymore.
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).
There's no point duplicating all the text that is already in the man
pages, and synchronizing them is a pain. Place a link to the github
generated pages instead.
Unfortunately, the anchor '#vo-opengl' does not work. Maybe github's
rst converter just sucks, as the actually generated HTML contains
links using that anchor too, but does not generate the anchor itself.
Too bad.
If the image is not writeable, the image actually has to be copied
beforehand. This was overlooked when converting the video chain to
reference counted images.
Fix a double free issue. This was overlooked when vf.c was changed to
free filter priv data automatically.
Use a different algorithm to generate the dithering matrix. This
looks much better than the previous ordered dither matrix with its
cross-hatch artifacts.
The matrix generation algorithm as well as its implementation was
contributed by Wessel Dankers aka Fruit. The code in dither.c is
his implementation, reformatted and with static global variables
removed by me.
The new matrix is uploaded as float texture - before this commit, it
was a normal integer fixed point matrix. This means dithering will
be disabled on systems without float textures.
The size of the dithering matrix can be configured, as the matrix is
generated at runtime. The generation of the matrix can take rather
long, and is already unacceptable with size 8. The default is at 6,
which takes about 100 ms on a Core2 Duo system with dither.c compiled
at -O2, which I consider just about acceptable.
The old ordered dithering is still available and can be selected by
putting the dither=ordered sub-option. The ordered dither matrix
generation code was moved to dither.c. This function was originally
written by Uoti Urpala.
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.
Notify the core of mouse movement events. The coordinates are converted to a
coordinate system with the origin in upper left corner, since Cocoa has it in
the lower left corner.
Use VOCTRL_CHECK_EVENTS instead. Change the remaining VOs to use it.
Only vo_sdl and vo_caca actually need this, and vo_null, vo_lavc, and
vo_image had stubs only.
Instead of having separate callbacks for each backend-handled feature
(like MPGLContext.fullscreen, MPGLContext.border, etc.), pass the
VOCTRL responsible for this directly to the backend. This allows
removing a bunch of callbacks, that currently must be set even for
optional/lesser features (like VOCTRL_BORDER).
This requires changes to all VOs using gl_common, as well as all
backends that support gl_common.
Also introduce VOCTRL_CHECK_EVENTS. vo.check_events is now optional.
VO backends can use VOCTRL_CHECK_EVENTS instead to implementing
check_events. This has the advantage that the event handling code in
VOs doesn't have to be duplicated if vo_control() is used.
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.
libavcodec generally shouldn't have this problem anymore (if libavcodec
ever had it). All other video decoders are gone. In any case, if this
commit actually causes regressions, these are libavcodec bugs and should
be fixed there instead.
This happens only if an option actually allocates memory (like strings).
Change filter API such that vf->priv is free'd by vf.c instead by the
filters. vf.c will free the option values as well.
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().
This tried to use ctx->pic (last decoded AVFrame) for the frame bounds.
However, if av_frame_unref() on the AVFrame is called, the function will
reset _all_ AVFrame fields, even those which are not involved with
memory management. As result, mpcodecs_config_vo() was called with 0
width/height, which made the function exit early, instead of
reconfiguring the filter chain.
Go back to using mpcodecs_config_vo() directly. (That's what it did
before this VDCTRL was originally introduced; the original reason for
it disappeared.)
vo_vdpau actually reads past the image allocation when displaying a
non-mod 2 420p image. The vdpau API specifies that VdpVideoSurfacePutBitsYCbCr()
requires a height that is a multiple of 4, and surface allocations are
automatically rounded.
So allocate video images with rounded height. libavutil does the same,
so images coming directly from the decoder or from libavfilter are no
problem. (libavutil does this alginment explicitly, not just because the
decoded image size is aligned to macroblocks.)
This changes the code so that it does the same as MPlayer, mplayer2
and mpv before ref-counted AVFrame. The problem is that get_buffer2
is called with aligned frame dimensions, while get_buffer didn't. This
breaks the mpv video frame size change detection.
Followup to 8df7127. This refines the condition for front ordering the
condition to account for minimized or hidden state where the window should go
to the front only as a consequnce of user interaction.
When using --fs `vo_cocoa_fullscreen` was called from the primary thread. This
occurred inside `vo_cocoa_config_window` which is scheduled for excution on the
primary thread with libdispatch).
All of this caused spam from NSRecursiveLock in standard output.
When going in and going out of full screen the player lost information on the
movableByWindowBackground behaviour. There were some hacks in place to fix it
but they were broken with the recent native fullscreen changes in 74c15ec6.
This commit removes the problem at the root and removes the hacks. The delegate
method `isMovableByWindowBackground` seems to be called after setFrame and
setPresentationOptions so change fs state in opts before that.
Fixes#83
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.
gl_video_resize_redraw() simply resizes and redraws (but without
invoking swapGlBuffers()). The VO is not involved in any way, so this
can simply be called from inside the mpgl lock from any thread.
Requires a minor refactor of the GL OSD code in order to redraw without
an OSD object.
To simplify things, we just assume that all OpenGL calls as well as
all calls into gl_video must be locked. Currently, also assume that
anything GUI related must be locked as well (stuff like VOCTRL_BORDER).
In its current state, this commit does nothing, but it will allow us to
move the Cocoa GUI out of the playloop, as well as possibly implementing
better framedropping.
Some OpenGL implementations on some platforms require that a context
is current only on one thread. For this reason, mpgl_lock() and
mpgl_unlock() take care of this as well for convenience.
Each backend that needs thread safety should provide it's own locking strategy
inside of `set_current`.
This fixes 2 bugs:
* Resizing very fast breaks the aspect of the window and the width and height
don't match with the video anymore
* Pressing 'f' for fullscreen very fast can overwrite the backup variables for
the previous width and height.
Also includes a better aspect calculation with fluid resizing.
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.)
This caused all formats with fewer than 8 bits per component marked as
little endian. (Normally, only some messed up packed RGB formats are
endian-specific in this case.)
This allows using the vdpau decoders with -vd without having to use
the -hwdec switch (basically like in mplayer).
Note that this way of selecting the hardware decoder is still
deprecated. libavcodec went away from adding special decoder entries
for hardware decoding, and instead makes use of the "hwaccel"
architecture, where hardware decoders use the same decoder names as
the software decoders. The old vdpau special decoders will probably
be deprecated and removed in the future.
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.
Useful for the j2k decoder.
Matrix taken from http://www.brucelindbloom.com/index.html?Eqn_RGB_XYZ_Matrix.html
(XYZ to sRGB, whitepoint D65)
Gamma conversion follows what libswscale does (2.6 in, 2.2 out).
If linear RGB is used internally for scaling, the gamma conversion
will be undone by setting the exponent to 1. Unfortunately, the two
gamma values don't compensate each others exactly (2.2 vs. 1/0.45=2.22...),
so output is a little bit incorrect in sRGB or color-managed mode. But
for now try hard to match libswscale output, which may or may not be
correct.
We only print an error message when POLLERR or POLLHUP occurrs, as the
something did go horribly wrong and the server will either deal with it or
crash.
Also add POLLOUT to the events.
Bump xkbcommon version and use the new xkb_keymap_from_buffer. This is more
secure, because the from_string expects a 0 terminated string, but this cannot
be guaranteed with mmap.
This commit remove a lot of linux specific code, like epoll. It also reduces
the complexity of the code. Instead of epoll we use poll which makes the
wayland backend more portable to other platforms.
This removes a good chunk of code trying to recreate key repeat.
Because the wayland protocol and xkbcommon don't have an interface for
auto-repeating pressed keys.
The code was attempting to get the ceiling of the double. Too bad NSSize has
floats inside of it and the int cast is nowhere to be seen. This caused
rounding errors by one pixel in the window size.
libavcodec decoder initialization failure caused a segfault, because it
wasn't properly reported back in init().
Also remove the return value from init_avctx(), which actually makes
things simpler. Instead, ctx->avctx can be checked to see whether
initialization was ok.
libavfilter changed the way a format list is passed to vf_format. Now
you have to separate formats with "|" instead of ":". If you use "|",
it prints an annoying message on every reinit:
[format @ 0x8bbaaa0]This syntax is deprecated. Use '|' to separate the list items.
...and it will probably stop working without warning at some point in
the future.
We need some very annoying ifdeffery to detect this case, because
libavfilter version numbers are just plain incompatible between Libav
and ffmpeg. There is no other way to detect this.
(Sometimes I wonder whether ffmpeg and especially Libav actually like
causing unnecessary pain for their users, and intentionally break stuff
in the most annoying way possible. Sigh...)
Resetting the filter graph helps dealing with filters which save state
between frames. This is important especially if they modify frame timing
or emit additional frames.
Unfortunately the libavfilter API doesn't have a way to do this
directly, so we have to use a dirty trick: we recreate the whole graph,
including format negotiation down and filter string parsing. ffplay does
this too. If libavfilter somehow decides to change output format or size
from what the first run in config() returned, mpv will explode. The same
applies to vf_next_query_format() return values (although this could be
mitigated, should it really happen).
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.
The VO warns by default that the nomanyfmts option should be used if
video display fails. This is almost completely useless, but people
keep asking what it means.
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.
This refactor makes the code easier to understand. Also corrects a bug that
caused the window to move to the left when the new size was bigger than the
visible frame.
It is now possible to show images and videos with alpha information correctly.
This was disalbed before, because there was a bug that made black parts of
videos also transparent.
