mythtv-vid: X11 Video output fixes branch

[danielk]

The point of the mythtv-vid branch is to fix some of the long standing deficiencies in the video output classes, for X11 in particular:

Filters

• Add OpenGL Lanczos filter
• Make setting up filters easier with a GUI
• Allow changing the filter set during playback, esp for deinterlacing.
• [DONE] Add support for OpenGL filtering for scaling and YV12->RGB conversion for display
• [DONE] Allow different filter sets for different video and display resolutions.
• [WONTFIX, see #1778] Auto-detect pull-up and apply ivtc filtering

Video output methods

• Implement per-frame video resolution switching
• Support XvMC for MPEG-2 NUV files (and enable transcode to MPEG-2)
• Key xrandr switching to video framerate
• [WONTFIX, VDPAU much more promising] Finish support for XvMC copy to pbuffer + opengl render

OSD rendering

• Fix video frame OSD blending for bob-deint (Filter vert freq in pause + when there is video<->screen framerate mismatch)
• Implement RGB OSD surfaces
• Implement X Composited OSD surfaces
• [DONE] Implement OpenGL OSD surfaces
• [DONE] Key preferred OSD and OSD fade to resolution and video output method.

[danielk]

(In [11431]) Refs #2287. Merges r10784:11430 from svn head to mythtv-vid.

[danielk]

(In [11495]) Refs #2287. Fix a copy-n-paste error that prevents us from using the selected OSD rendering method in mythtv-vid tree.

[danielk]

(In [11497]) Refs #2287. Merges -r11430:11496 in svn head to mythtv-vid.

[danielk]

(In [11498]) Refs #2287. Changes VDP to tell you why a particular decoder or rendering method was rejected.

[danielk]

(In [11504]) Refs #2123, refs #2287. Allows deleting display profile entries in the Playback Profiles Editor in mythtv-vid.

[danielk]

(In [11505]) Refs #2123, refs #2287. Updates the display profile entry labels after you edit an entry in the Playback Profiles Editor in mythtv-vid.

[danielk]

(In [11506]) Refs #2123, refs #2287. Allows adding display profile entries in the Playback Profiles Editor in mythtv-vid.

[danielk]

(In [11509]) Refs #2123, refs #2287. Allows adding and deleting display profile groups in the Playback Profiles Editor in mythtv-vid.

There are still bugs in this code, please don't open new tickets on them until #2123 is closed; mostly the bugs have to do with screens not repainting appropriately or initial values being screewy, so they can be worked around.

[danielk]

(In [11608]) Refs #2287. Merges -r11496:11607 in svn head to mythtv-vid.

[danielk]

(In [11661]) Refs #2287. Fixes #2634. Workaround for HD streams that report their resolution as 1440x1088 when they are really 1440x1080 (with patch from Mark Kendall).

[danielk]

(In [11674]) Refs #2123. Refs #2287. Fixes bug introduced in [11509] which broke recording profiles with patch from skamithi.

[danielk]

(In [11806]) Refs #2287. Merges -r11607:11700 in svn head to mythtv-vid

[danielk]

(In [11807]) Refs #2287. Merges -r11700:11806 in svn head to mythtv-vid.

This breaks some of the UI, this will be addressed in a separate changeset.

[danielk]

(In [11810]) Refs #2287. Fixes some globalsettings.{cpp,h} problems in mythtv-vid.

[danielk]

(In [11811]) Refs #2287. Fixes various problems with the recent merge from head to svn head.

The ConfigurationDialog needed a Storage implementation, so that it's children could override load() and save(). Also various configuration dialogs were crowded due to the changes in the settings from the last merge. Finally the triggered configuration group had some problems with deleting panes after the merge.

There remain problems with the PlaybackProfileConfig loading and saving, especially with respect to new configs.

[danielk]

(In [12221]) Refs #2287. Merges -r11806:12220 in svn head to mythtv-vid.

[danielk]

(In [12257]) Refs #2287. Merges -r12220:12256 from svn head to mythtv-vid.

[danielk]

(In [12266]) Refs #2287. Rearranges some configuration options to fit on the screen @640x480 in the mythtv-vid branch.

[danielk]

(In [12676]) Refs #2287. Merges -r12256:12675 from svn head to mythtv-vid.

[danielk]

(In [12688]) Refs #2287. Adds OSD fade configurable in the video profiles.

Instead of hard coding the disabling of the OSD fade for XvMC and HDTV this changeset makes this a controlable properlty of the video display profile, so you can enable fading when you have a fast GPU or CPU.

If you were already using the mythtv-vid branch, I recommend dropping the displayprofilegroups and displayprofiles tables. Otherwise this property will not be configurable.

[danielk]

(In [12693]) Refs #2287. Refs #2649. Adds Mark Kendall's OpenGL video rendering to mythtv-vid tree.

I've made this default to being included in ./configure. You just need to select opengl as a video renderer for one of your display profiles to use this. Here it looks pretty rotten with SD material, but looks quite good with HD material.

[danielk]

(In [12695]) Refs #2287. Enables compilation of XvMC w/OpenGL OSD by default when '-xvmc-lib=XvMCNVIDIA' is used to explicitly link to the nVidia XvMC library.

[danielk]

(In [12731]) Refs #2287. Makes renderer fallback work with video display profiles when initialization of the preferred video renderer fails.

[danielk]

(In [12732]) Refs #2287. Refs #2649. Uses utils-opengl functions for OpenGL VSync.

[danielk]

(In [12734]) Refs #2287. Allow ivtv to be disabled in mythtv-vid branch.

[Mark.Buechler@…]

Changing channels to new tuner ignores channel and defaults to stored cardinput.startchan instead with no indication of selected channel in log nor any lines stating it being bad. Tuning from there to desired channel works. I changed from channel 244 to 122 and mythbackend selected the stored channel 9429:

2007-02-05 18:10:23.615 MainServer::HandleAnnounce Playback 2007-02-05 18:10:23.617 adding: pc4 as a client (events: 0) 2007-02-05 18:10:23.635 TVRec(5): Changing from None to WatchingLiveTV 2007-02-05 18:10:23.641 TVRec(5): HW Tuner: 5->5 2007-02-05 18:10:23.646 DVBChan(2): Opening DVB channel 2007-02-05 18:10:23.648 DVBChan(2): SetChannelByString(9429): 2007-02-05 18:10:23.653 ChannelBase(5): Failed to find channel(9429) on current input (DiSEqC v1.3 Input 11) of card (5). 2007-02-05 18:10:23.657 ChannelBase(5): Found channel(9429) on another input (DiSEqC v1.3 Input 5) of card (5).

[danielk]

(In [12736]) Refs #2287. Works around a bug in the nVidia drivers which can cause a SEGFAULT when we call XCloseDisplay().

[danielk]

(In [12737]) Refs #2287. Implements video renderer fallback in mythtv-vid branch.

This allows you to compile in multiple conflicting video output methods, such as Linux framebuffer support and X11 support and select the preferred one at runtime based on your video display profile.

[danielk]

(In [12738]) Refs #2287. Fixes a bug in fallback code in [12737] wrt xvmc surfaces.

[John Ball]

Hi, It looks like 12737 isn't happy when ivtv drivers aren't installed:

NuppelVideoPlayer.cpp: In member function `void NuppelVideoPlayer::SetDecoder(DecoderBase*)':
NuppelVideoPlayer.cpp:6538: error: `IvtvDecoder' was not declared in this scope
NuppelVideoPlayer.cpp:6538: error: `ivtv' was not declared in this scope
NuppelVideoPlayer.cpp:6538: error: `IvtvDecoder' is not a type
NuppelVideoPlayer.cpp:6538: error: expected `>' before '*' token
NuppelVideoPlayer.cpp:6538: error: expected `(' before '*' token
NuppelVideoPlayer.cpp:6538: error: expected primary-expression before '>' token
NuppelVideoPlayer.cpp:6538: error: expected `)' before ';' token
NuppelVideoPlayer.cpp:6538: warning: unused variable 'IvtvDecoder'
make[2]: *** [NuppelVideoPlayer.o] Error 1
make[2]: Leaving directory `/home/mythtv/src/testing-mythvid/mythtv-vid/libs/libmythtv'
make[1]: *** [sub-libmythtv] Error 2
make[1]: Leaving directory `/home/mythtv/src/testing-mythvid/mythtv-vid/libs'
make: *** [sub-libs] Error 2

I'm using --disable-ivtv. Thanks!

-John

[danielk]

(In [12743]) Refs #2287. Fixes two compiling problems, one for disabling ivtv support, the other for gcc 3.x compilation.

[danielk]

(In [12752]) Refs #2287. Adds a toString() function to MythCodedID's in mythtv-vid

[danielk]

(In [12753]) Refs #2287. Fixes opengl->other video renderer swiching in VideoOutputXv::InputChanged().

[danielk]

(In [12758]) Refs #2287. Reworks some ifdef's and silences a couple compile time warnings. This allows all the various video output methods that work on Linux to be enabled/disabled invidually (unless one depends on another, XVideo support is needed for any output method in VideoOutputXv).

[danielk]

(In [12761]) Refs #2287. Cleanup of DirectFB code.

This needs more testing. I'm using an emulated framebuffer...

[danielk]

(In [12770]) Refs #843. Refs #2287. This temporarily reverts out the accelerated PiP.

This is causing various problems with PiP. I'm going to make some simplifying changes to tv_play video player handling then apply this back in smaller chunks which should be easier to debug.

[danielk]

(In [12949]) Refs #2287. Merges r12675:12948 from svn head to mythtv-vid.

[danielk]

(In [13287]) Refs #2287. Fixes #3044. Fixes #2649. OpenGL video renderer improvements and fixes.

This improves the speed of the YUV->RGB conversion by accessing the texels with greater coherency. This fixes removes a doubled call to the resize filter. This avoids a segfault by leaking some memory (temporary fix).

[danielk]

(In [13433]) Refs #2287. Merges r12948:13432 from svn head to mythtv-vid.

[danielk]

(In [13545]) Refs #2287. Merges r13433:13544 from svn head to mythtv-vid

[danielk]

(In [13548]) Refs #2287. Refs #2649. Fixes a bug in XvMC-OpenGL introduced by [12708].

We changed to using glTexSubImage2D() to fill the OSD texture, but we're losing the alpha so it covers up the video. This changes back to using glTexImage2D().

[danielk]

(In [13549]) Refs #2287. Fixes bug in XvMC-OpenGL pause handling.

The XvMC-OpenGL video renderer stopped updatating the OSD once the video was paused because PrepareFrame was doing the compositing step. This changeset moves the compositing step to the Show() method so that the OSD is still updated when the video is paused when using the XvMC-OpenGL video renderer.

[danielk]

(In [13551]) Refs #2287. Compilation fix for non-XvMC-OpenGL. (Missing method signature change in [13549] for #ifdef'd code, spotted by Stanley Kamithi.)

[danielk]

(In [13552]) Refs #2287. Fix for OpenGL video render's DVD still frame display. The problem was discovered by Stanley Kamithi and fixed by Mark Kendall.

[danielk]

(In [13767]) Refs #2287. Fixes pausing using OpenGL video renderer when the display window size is not divisible by 4 and 2 for the width and height, resp.

We need to adhere to I420 since we are rendiring the OSD in that color space.

[danielk]

(In [13769]) Refs #2287. Workaround for an apparent nvidia bug that caused blank screens after video rendering method changes.

The problem was noticed by Stanley Kamithi and worked around by Mark Kendal. The video playback window would be blank when switching from OpenGL to XvMC rendering with 8776 and crash when using the 9755 drivers, This may have something to do with the change to using the 3D pipeline to emulate XVideo in the newer drivers. The workaround was simply to unmap and map the subwindow with the OpenGL context instead of deleting and recreating it as we switch between rendering methods. The window and context are still properly deleted when we delete VideoOutputXv at playback shutdown.

[danielk]

(In [13770]) Refs #2287. Merges r13544:13769 from svn trunk to mythtv-vid.

[danielk]

(In [13771]) Refs #2287. Avoids a division by zero for very small video output windows in VideoOutputXv::GetVisibleOSDBounds().

[matt.doran@…]

As requested by danielk, when enabling the hardware deinterlacing I see errors in the log like this:

2007-06-28 23:45:15.560 VDP: GetFilteredDeint() : -> 'opengllinearblend' 2007-06-28 23:45:15.564 FilterManager: failed to load filter 'opengllinearblend', no such filter exists 2007-06-28 23:45:15.564 Couldn't load deinterlace filter opengllinearblend 2007-06-28 23:45:15.564 Using deinterlace method 2007-06-28 23:45:15.564 SetVideoRenderer(opengl)

This might be expected, because it's doing the deinterlacing in the hardware rather than the software filter ... but it's a bit misleading and will probably cause some confusion.

[danielk]

(In [14057]) Refs #2287. Some tweaks to videodisplayprofile. This removes the 'last_video_renderer' check in SetVideoRenderer(), if LoadBestPreferences() has been called since SetVideoRenderer() was last called the preferences for the deinterlacers, osd renderers, etc may no longer be compatable with the video renderer, so we need to recheck this.

This also adds some debugging output and names the quartz-accel OSD renderer.

[danielk]

(In [14058]) Refs #2287, Moves myth2av_codecid from videoout_xv.cpp to mythcodecid.cpp, and adds DVDV codecs to mythcodecid. This is in preperation for using myth2av_codecid in videoout_quartz.cpp

[danielk]

(In [14059]) Refs #2287. Adds link line to OSX build needed for Objective C code to link.

[danielk]

(In [14060]) Refs #2287. Fixes a regression in the ability to enable deinterlacing.

Basically the SetScanType() just ignored any attempts to enable deinterlacing because we removed the m_DeintSetting from NuppelVidepPlayer and didn't replace it with anything now that the video renders track the deinterlacing settings.

[danielk]

Patch to add OSX DVDV decoding (not functionning).

[danielk]

(In [14062]) Refs #2287. Removes hard coded "bobdeint" for XvMC deinterlacing.

The deinterlacing method can now be controlled for XvMC again, you have two options "oneline" and "bobdeint".

[danielk]

(In [14063]) Refs #2287. #1279. This applies an updated version of DVDV code to libavcodec.

John Dagliesh wrote the original DVDV code and Nigel did the original merge, see #1279, this just updates the merge to work with the latest changes in how ffmpeg handles codecs and commits the mpeg12.c and dvdv.c portions of the patch.

you need to rerun ./configure before recompiling (to pick up the new codec define.)

[danielk]

(In [14064]) Refs #2287. Changes ./configure to enable XvMC OpenGL support by default if you explicitly link to libXvMCNVIDIA. It is still only used if you enable it in your display profiles.

[danielk]

(In [14065]) Refs #2287. Reverts everything but ./configure change in [14064], I accidentally committed a bunch of extra code from another patch.

[danielk]

(In [14068]) Refs #2287. This makes 3 changes to the VideoOutput::InputChanged() and a complimentary change to VideoOutput::Create().

The main change is to add a new param "codec_private". The OS X DVDV playback renderer needs access to the DVDV decoder which is created in AvFormatDecoder. This allows this data to be passed to VideoOutputQuartz in a generic way which we could reuse in the future without changing the InputChanged() signature again.

The two smaller changes were things I've wanted in the API for a while, but were too trivial to make the API change for alone. The changes are to: (1) return a ok/failed status from InputChanged(). Changing the resolution of a video can fail if the resolution exceeds what the hardware is capable of, as in #3680. In [14009], I was able to check the error status after calling InputChanged() to avoid the segfault. But actually returning this information is much nicer API wise. (2) The other is to use a QSize for the dimensions of the video, this avoids us needing to convert from QSize to ints and back to QSize when calling InputChanged().

Finally, VideoOutput::Create() just had the "codec_private" paramater added so that we can pass the initial information in when creating a VideoOutput instance. This wasn't added to VideoOutput::Init() since we can just pass the codec_private into the constructors that need it and keep the changeset much smaller.

[danielk]

(In [14069]) Refs #2287. Restores ApproveDeintFilter() check in VideoOutput::SetupDeinterlace().

ApproveDeintFilter() got in the way of testing the display profile code, but is still needed when no matching profile is found for a particular video.

[danielk]

(In [14070]) Refs #2287. Refs #3523. Various minor fixes to VideoOutputQuartz.

• Changes QuartzData from a struct to a class, so we can properly initialize all it's variables without an init function.
• Sets data pointer to NULL after deleting the class for easier debugging.
• Calls SetVideoRender() in the ctor and CreateQuartzBuffers(), the latter is needed in case there is no quartz video profile, otherwise the VideoOutputBase::Init() call may reset the renderer.
• Sets display_dim and display_aspect properly. This is a prerequisite to addressing #3523,
• Removes incorrect code in VideoOutputQuartz::GetAllowedRenderers() for enabling DVDV rendering.
• Tidys up some VERBOSE macros.

[freedenizen@…]

I noticed that Vertical Scaling doesn't appear to be honored by the opengl renderer. I haven't had a chance to test the other overscan options to see if they work or not.

[danielk]

(In [14094]) Refs #2287. Adds a profile for OSX quartz-blit rendering to CPU++ video profile group.

[danielk]

(In [14095]) Refs #2287. Updates DVDV hooks for mythtv-vid branch.

This is disabled by default and Nigel is the only person to have gotten this change to work partially (i.e. with a couple bad crashes.) But it's there for you to try using the "--enable-dvdv" ./configure flag..

[danielk]

(In [14096]) Refs #2287. Reformats videoout_dvdv.{h,cpp} in MythTV convention and comments out NSAutoreleasePool code that was causing a segfault. (The NSAutoreleasePool will be dealt with in a seperate changeset.)

[danielk]

(In [14098]) Refs #2287. Creates and destroyes the memory allocator needed by Cocoa.

The OS X Cocoa libraries assume that you have a garbage collector allocated and will leak memory if you don't. However it needs a specific garbage collector for each thread which creates some complications. This creates and destroyes the garbage collector needed to handle DVDV stuff.

We really should allocate a garbage collector for each thread in MythTV on the OS X platform, but this addresses the only major memory leak this Cocoa requirement appears to cause. And addresses all the additional memory leaks DVDV could have added to the mix.

[danielk]

(In [14113]) Refs #2287. Disables video preview in EPG when you use OpenGL video renderer.

This is a patch from Mark Kendal, The OpenGL rendering method doesn't support video preview in the EPG so we don't want to waste the screen real estate on it.

[danielk]

(In [14115]) Refs #2287. This solves the remaining problems with OS X garbage collection.

Written by me, tested by Nigel.

[danielk]

(In [14254]) Refs #2287. Makes libav always use custom allocator and flags for DVDV decoding and adds an error message when any direct rendered codec doesn't use a MythTV allocator (and uses MythTV's default direct rending friendly allocator in this case).

[danielk]

(In [14255]) Refs #2287. Sets the myth binary revision and protocol revisions to numbers that don't conflict with trunk.

This requires plugins to be rebuilt. Basically some of the reports in the mythtv-vid testing threads were caused by not recompiling plugins for mythtv-vid or using a mythtv-vid frontend with a non-mythtv-vid backend this should prevent this type of problem for testers.

[danielk]

(In [14284]) Refs #2287. Rearranges a couple config pages with an odd tab order, inspired by comment by Allan Stirling.

[danielk]

(In [14285]) Refs #2287. Adds GL_FRAMEBUFFER_INCOMPLETE_DUPLICATE_ATTACHMENT_EXT define if needed for debian header compatibility. Problem reported by Doug Larrick.

[danielk]

(In [14286]) Refs #2287. Merges r14097:14285 from svn head to mythtv-vid.

[skamithi]

(In [14361]) Refs #2287. fix an indentation problem, and fix a problem where opengl renderer doesn't use software conversion if a resize filter is not found.

[skamithi]

(In [14370]) Refs #2287. similiar to [14361]. fix the issue where the opengl2 osd doesn't fail to initialize if a resize filter cannot be found.

[danielk]

(In [14403]) Refs #2287. Merges r14285:14401 from svn trunk to mythtv-vid.

[danielk]

(In [14405]) Refs #2287. Merges mythtv-vid to head.

You must rerun ./configure before comiling and recompile and install the plugins after installing this.

Unlike the mythtv-vid branch, compiling OpenGL rendering is disabled due to some instability issues.

There are a few video related things in here:

• Adds Video Profiles and a Video Profiles Editor to Playback config. This allows you to use different rendering methods and deinterlacer options for different video resolutions.
• Adds an accelerated playback "DVDV" for OS X
• This enables compiling XvMC chromakey rendering by default
• This enables compiling DirectFB rendering by default, and adds support for the newish DirectFB 1.0
• This enables compiling PVR-350 rendering by default
• This splits OSD configuration out from the general Playback config. (A separate commit will add this to the external themes.)

Please report any problems on the #2287 mythtv-vid ticket.

[danielk]

(In [14408]) Refs #2287. Merges r14401:14407 from svn trunk to mythtv-vid.

[otto at kolsi dot fi]

Replying to danielk:

You must rerun ./configure before comiling and recompile and install the plugins after installing this.

Having problems compiling trunk after merge. Same problem occurs with my two machines (Athlon & PIV Celeron). I have done "make distclean" and re-run configure. Also compiled trunk just minutes before merge and that version compiled fine.

g++ -c -pipe -march=i686 -Wall -W -fomit-frame-pointer -O3 -Wall -Wno-switch -Wpointer-arith -Wredundant-decls -Wno-non-virtual-dtor -D__STDC_CONSTANT_MACROS -fomit-frame-pointer -I/usr/include/freetype2 -D_REENTRANT -DPIC -fPIC  -DMMX -Di386 -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64 -DPREFIX=\"/usr/local\" -DLIBDIR=\"/usr/local/lib\" -D_LARGEFILE_SOURCE -DUSING_H264TOOLS -DUSING_X11 -DUSING_XV -DUSING_XVMC -DUSING_XVMCW -DUSING_OPENGL_VSYNC -DUSING_FRONTEND -DUSING_DVB -DUSING_BACKEND -DQT_NO_DEBUG -DQT_THREAD_SUPPORT -DQT_SHARED -I/usr/lib/qt-3.3/mkspecs/default -I. -I/usr/local/include -I/usr/include -I../.. -I.. -I../libmyth -I../libavcodec -I../libavutil -I../libmythmpeg2 -Idvbdev -Impeg -Iiptv -I../libmythlivemedia/BasicUsageEnvironment/include -I../libmythlivemedia/groupsock/include -I../libmythlivemedia/liveMedia/include -I../libmythlivemedia/UsageEnvironment/include -I/usr/lib/qt-3.3/include -o NuppelVideoPlayer.o NuppelVideoPlayer.cpp
g++ -c -pipe -march=i686 -Wall -W -fomit-frame-pointer -O3 -Wall -Wno-switch -Wpointer-arith -Wredundant-decls -Wno-non-virtual-dtor -D__STDC_CONSTANT_MACROS -fomit-frame-pointer -I/usr/include/freetype2 -D_REENTRANT -DPIC -fPIC  -DMMX -Di386 -D_GNU_SOURCE -D_FILE_OFFSET_BITS=64 -DPREFIX=\"/usr/local\" -DLIBDIR=\"/usr/local/lib\" -D_LARGEFILE_SOURCE -DUSING_H264TOOLS -DUSING_X11 -DUSING_XV -DUSING_XVMC -DUSING_XVMCW -DUSING_OPENGL_VSYNC -DUSING_FRONTEND -DUSING_DVB -DUSING_BACKEND -DQT_NO_DEBUG -DQT_THREAD_SUPPORT -DQT_SHARED -I/usr/lib/qt-3.3/mkspecs/default -I. -I/usr/local/include -I/usr/include -I../.. -I.. -I../libmyth -I../libavcodec -I../libavutil -I../libmythmpeg2 -Idvbdev -Impeg -Iiptv -I../libmythlivemedia/BasicUsageEnvironment/include -I../libmythlivemedia/groupsock/include -I../libmythlivemedia/liveMedia/include -I../libmythlivemedia/UsageEnvironment/include -I/usr/lib/qt-3.3/include -o vsync.o vsync.cpp
vsync.cpp: In destructor ‘virtual OpenGLVideoSync::~OpenGLVideoSync()’:
vsync.cpp:456: error: ‘glXDestroyContext’ was not declared in this scope
vsync.cpp: In member function ‘virtual bool OpenGLVideoSync::TryInit()’:
vsync.cpp:489: error: ‘init_opengl’ was not declared in this scope
vsync.cpp:489: error: ‘get_glx_version’ was not declared in this scope
vsync.cpp:497: error: ‘glXQueryExtensionsString’ was not declared in this scope
vsync.cpp:499: error: ‘has_glx_video_sync_support’ was not declared in this scope
vsync.cpp:505: error: ‘GLX_RGBA’ was not declared in this scope
vsync.cpp:506: error: ‘GLX_RED_SIZE’ was not declared in this scope
vsync.cpp:507: error: ‘GLX_GREEN_SIZE’ was not declared in this scope
vsync.cpp:508: error: ‘GLX_BLUE_SIZE’ was not declared in this scope
vsync.cpp:514: error: ‘glXChooseVisual’ was not declared in this scope
vsync.cpp:544: error: ‘GL_TRUE’ was not declared in this scope
vsync.cpp:544: error: ‘glXCreateContext’ was not declared in this scope
vsync.cpp:552: error: ‘glXMakeCurrent’ was not declared in this scope
vsync.cpp:556: error: ‘gMythGLXGetVideoSyncSGI’ was not declared in this scope
vsync.cpp:562: error: ‘GLX_BAD_CONTEXT’ was not declared in this scope
vsync.cpp: In function ‘bool checkGLSyncError(const QString&, int)’:
vsync.cpp:595: error: ‘GLX_BAD_CONTEXT’ was not declared in this scope
vsync.cpp:598: error: ‘GLX_BAD_VALUE’ was not declared in this scope
vsync.cpp: In member function ‘virtual void OpenGLVideoSync::Start()’:
vsync.cpp:627: error: ‘glXMakeCurrent’ was not declared in this scope
vsync.cpp:637: error: ‘gMythGLXGetVideoSyncSGI’ was not declared in this scope
vsync.cpp:639: error: ‘gMythGLXWaitVideoSyncSGI’ was not declared in this scope
vsync.cpp:642: error: ‘glXMakeCurrent’ was not declared in this scope
vsync.cpp: In member function ‘virtual void OpenGLVideoSync::WaitForFrame(int)’:
vsync.cpp:661: error: ‘glXMakeCurrent’ was not declared in this scope
vsync.cpp:664: error: ‘gMythGLXGetVideoSyncSGI’ was not declared in this scope
vsync.cpp:670: error: ‘gMythGLXWaitVideoSyncSGI’ was not declared in this scope
vsync.cpp:679: error: ‘gMythGLXWaitVideoSyncSGI’ was not declared in this scope
vsync.cpp:684: error: ‘glXMakeCurrent’ was not declared in this scope
make[2]: *** [vsync.o] Error 1
make[2]: Leaving directory `/mnt/myth/mythtv/mythtv/libs/libmythtv'
make[1]: *** [sub-libmythtv] Error 2
make[1]: Leaving directory `/mnt/myth/mythtv/mythtv/libs'
make: *** [sub-libs] Error 2

[danielk]

(In [14410]) Refs #2287. Fixes a compilation problem when opengl is needed and neither OpenGL video or XvMC OpenGL are enabled.

[danielk]

(In [14416]) Refs #2287. Adds new TV_SETTINGS_PLAYBACK_OSD buttons to myththemes themes.

[skamithi]

(In [14418]) Refs #2287. check if openGL filter contains frameBuffers. prevent crash when using software colour conversion.

[danielk]

(In [14421]) Refs #2287. Allows compiling with OpenGL headers for hardware/software which does not support the GL_EXT_framebuffer_object OpenGL extension.

[danielk]

(In [14423]) Refs #2287. Fixes issue with enabling xvmc-vld*, If OpenGL was enabled at all it would be disabled due to missing brackets. This also makes the whole --enable-xvmc* infrastructure a little more foolproof.

[laga@…]

Hi.

• There's a typo in ./configure --help output:

    --enable-opengl-video   disable opengl based video display
  

should probably read "enable opengl based video display".

• When editing the video display profile, I can select osd renderers/decoders which were not actually compiled in. Of course, it falls back to sane defaults and lets me watch TV, but it might be better not to display choices which were not compiled in. Just a thought.

Michael

[laga@…]

Hi,

I'm using an interlaced PAL modeline to drive my TV:

Modeline "720x576pali" 13.88 720 726 794 880 576 580 583 625 -hsync -vsync interlace

My video/osd render method is opengl. When i'm trying to use the normal (software) bodeint, i get the following messages:

2007-09-07 15:44:48.131 VDP: GetFilteredDeint() : opengl -> 'bobdeint'
2007-09-07 15:44:48.132 VDP: GetFilteredDeint() : opengl -> 'bobdeint'
2007-09-07 15:44:48.133 Failed to approve 'bobdeint' deinterlacer
2007-09-07 15:44:48.134 Couldn't load deinterlace filter

However, the hardware bobdeint works:

2007-09-07 15:42:36.081 VDP: GetFilteredDeint() : opengl -> 'openglbobdeint'
2007-09-07 15:42:36.081 GLVid: Creating openglbobdeint filter.
2007-09-07 15:42:36.082 GLVid: Created fragment program openglbobdeint.
2007-09-07 15:42:36.091 GLCtx: Created frame buffer object (720x576).
2007-09-07 15:42:36.092 GLVid: Turning off deinterlacing.
2007-09-07 15:42:36.092 VideoOutputXv: Using deinterlace method openglbobdeint
2007-09-07 15:42:36.092 GLVid: Turning on deinterlacing.

Prior to the mythtv-vid merge, bobdeint would never work for me because my refresh rate was too low. I had to use the patch from #2903 to enable bobdeint (without a deinterlacer, I would get interlace artefacts/stuttering despite my video resolution matching my display resolution - same happens to a lot of other people, too). Now I'm kinda confused by this discrepancy - is it intentional that openglbobdeint works while bobdeint does not?

I should note that I'm using a Twinview setup. CRT-0 is a 19" CRT monitor running at 1280x102, CRT-1 is my TV running said modeline.

[danielk]

(In [14447]) Refs #2287. Fixes help text error in ./configure output, We said an option disabled opengl-video when it enabled it. Reported by laga at laga ath cx.

[danielk]

(In [14451]) Refs #2287. Merges r14407:14450 from svn trunk to mythtv-vid.

[danielk]

(In [14452]) Refs #2287. Adds support for GLX 1.2 for OpenGL renderer (previously required GLX 1.3 or better).

This was written by Stanley Kamithi with small modifications by me. This allows Intel and ATI cards to take advantage of the Mark Kendall's OpenGL rendering method.

[danielk]

(In [14453]) Refs #2287. Two bug fixes from Stanley Kamithi for software filtering config.

1st/ we called VideoDisplayProfile::IsFilterAllowed() with the filter value rather than the correct renderer. 2nd/ The editable ComboBox is a little buggy so we're returning to a LineEdit for now.

[danielk]

(In [14454]) Refs #2287. Applies an OpenGL rendering debugging patch from Mark Kendall.

[danielk]

(In [14455]) Refs #2287. Adds another OpenGL deinterlacer from Mark Kendall.

[danielk]

(In [14456]) Refs #2287. Applies patch by Mark Kendall to properly handle overlays when overscan is in use with the OpenGL video rendering method.

[danielk]

(In [14457]) Refs #2287. Applies portion of OpenGL video rendering locking patch from Mark Kendall + some other locking issues I noticed when reviewing the changes.

[skamithi]

(In [14458]) Refs #2287. fix compile error with xvmctextures.cpp because of [14452]

[danielk]

(In [14459]) Fixes #3486. Refs #2287. Applies patch from Ian Armstrong which adds support for latest IVTV ioctl's in for PVR-350 playback.

[danielk]

(In [14477]) Refs #2287. Fixes #1842. Fixes #3943. Fixes #3479. Merges mythtv-vid to svn trunk.

After the merge last week, I applied some patches that were waiting in the queue for mythtv-vid.

• Fixes some issues with VIA CN400 drivers not displaying 1080i frames.
• Adds a workaround to for crashes due to bugs in the Intel and VIA OpenGL drivers.
• Allows OpenGL rendering to work with GLX 1.2 for ATI and Intel cards.
• Seperates preconfigured zoom and aspect ratio override modes, "W" is still the binding for zoom, and "Ctrl-W" now overrides the aspect ratio specified in a video file.
• Adds a new OpenGL video renderer output filter from Mark Kendall.
• Reworks some of the OpenGL video renderer locking for greater stability.

[skamithi]

(In [14479]) Refs #2287. when using the chromakey osd check for the first available xv port that supports the XV_COLORKEY attribute. some xv implementations, such as Intel drivers (v2.1), list textured video adaptors before video overlay adapters.

[danielk]

(In [14490]) Refs #2287. Fix from Mark Kendall for a software bobdeint with OpenGL output.

The problem was originally reported by Tino Keitel as a split screen when using Intel i810 graphics drivers.

[skamithi]

(In [14512]) Refs #2287. fix opengl pip positioning. Patch from Mark Kendall.

[skamithi]

(In [14568]) Refs #2287. merges r14452:14567 from svn head to mythtv-vid.

[skamithi]

(In [14569]) Refs #2287 commit [14490] to mythtv-vid. didn't get added during last merge from svn head to mythtv-vid.

[skamithi]

(In [14571]) Refs #2287. fix opengl video resizing. patch prerequisite for allowing embedding of opengl window in program guide or playbackbox. Patch from Mark Kendall.

[skamithi]

(In [14580]) Refs #843. Refs #2287. merge mythtv-pip to mythtv-vid.

[skamithi]

(In [14588]) Refs #2287. mythtv-vid: fix debug compilation. broken after the last merge. comm flagging is also broken after the merge from mythtv-pip. will fix this soon.

[danielk]

(In [14590]) Refs #2287. Fix for OpenGL Renderer picture controls, these weren't working for pre-recorded material because the OpenGL filter relied on use_picture_controls to be set before InitVideoBuffers is called; but they aren't set until after InitVideoBuffers(); with this change it is set based on the DB setting alone before InitVideoBuffers().

[skamithi]

(In [14591]) Refs #2287. Refs #843 fix use of the player context in mythcommflag.

[danielk]

(In [14599]) Refs #2287. Return a different mpeg version for CODEC_ID_H264 than CODEC_ID_MPEG4.

In ffmpeg, MPEG4 refers to MPEG-4 Part 2 video, while H264 refers to MPEG-4 Part 10 video, and completely different video decoder is used for these two types of MPEG-4 video streams so we need to keep track of this for ffmpeg's sake, using the pseudo MPEG version 5 for MPEG-4 part 10 allows us to do this.

[skamithi]

(In [14618]) Refs #2287. merge r14574:14612 from trunk to mythtv-vid

[skamithi]

(In [14661]) Refs #2287. merge r14568:r14659 from trunk to mythtv-vid.

[skamithi]

(In [14678]) Refs #843. Refs #2287. change method for embedding TV video in guidegrid and playbackbox. Allows opengl/xv/xvmc-blit rendered TV windows to be embedded. xvmc-opengl doesn't work yet cause xvmc-opengl playback is broken. talking to danielk, there are some thread safety issues with embedded video playback to deal with.

[skamithi]

(In [14696]) Refs #843 Refs #2287. mythtv-vid: fix bug with playercontext code that was causing hang during live tv window destruction.

[skamithi]

(In [14709]) Refs #2287. merge r14613:14707 from trunk to mythtv-vid

[skamithi]

(In [14710]) Refs #843 Refs #2287. remove playercontext code in mythcommflag. causing intermittent commflagging failures. may fix and re-add later.

[skamithi]

(In [14718]) Refs #2287. fix some merge problems found after the last update from trunk.

[skamithi]

(In [14775]) Refs #2287 Refs #843. mythtv-vid: playercontext code: fix crash while watching live tv.

[danielk]

(In [14778]) Refs #2287. Merges r14707:14777 from trunk to mythtv-vid

[skamithi]

(In [14832]) Refs #2287. merges 14779:14831 from trunk to mythtv-vid.

[skamithi]

(In [15159]) Refs #2287. r14961:15157 from trunk to mythtv-vid

[dennisoberhoff@…]

Replying to danielk:

(In [13769]) Refs #2287. Workaround for an apparent nvidia bug that caused blank screens after video rendering method changes.

The problem was noticed by Stanley Kamithi and worked around by Mark Kendal. The video playback window would be blank when switching from OpenGL to XvMC rendering with 8776 and crash when using the 9755 drivers, This may have something to do with the change to using the 3D pipeline to emulate XVideo in the newer drivers. The workaround was simply to unmap and map the subwindow with the OpenGL context instead of deleting and recreating it as we switch between rendering methods. The window and context are still properly deleted when we delete VideoOutputXv at playback shutdown.

This Bug, or something similar, seems to be appearing with the latest (beta) Drivers from Nvidia. Closing and reopening the Internal Videoplayer causes an blank Screen.

[skamithi]

(In [15451]) Refs #2287 r15158:15450 from trunk to mythtv-vid

[skamithi]

(In [15452]) Refs #2287 Refs #3797. add gray letterboxing support for opengl renderer.

[skamithi]

(In [15453]) Refs #2287. prevent a crash that occurs when you run opengl renderer with more then 2 openglvideo instances and then you start and stop the tv player several times. e.g opengl with opengl osd or opengl with pip. it crashed with a double link list or Xorg crashes. patch makes openglvideo::initopengl execute one the initial openglvideo instance instead of once per openglvideo instance.

[skamithi]

(In [15455]) Refs #2287. implement opengl vsync using openglcontext code.

[skamithi]

(In [15458]) Refs #2287. fix merge error regarding [15457]

[skamithi]

(In [15507]) Refs #2287.prev attempt to implement opengl vsync using openglcontext, got opengl vsync to share the same context as opengl renderer if present. this caused a problem if changing from opengl->xv renderer within the same TV playback window. so created a separate OpenGLcontext for vsync.

[skamithi]

(In [15530]) Refs #2287. pause the NVP video loop when embedding the video. helps prevent some crashes during the video resize process.

[skamithi]

(In [15532]) Refs #2287 Refs #843. disable opengl vsync functions when the video is embedded. helps prevent some hangs during video resizing when opengl vsync is enabled.

[skamithi]

(In [15601]) Refs #843. Refs #2287. last merge from mythtv-pip to mythtv-vid before deleting mythtv-pip.

[skamithi]

(In [15626]) Refs #2287. mythtv-vid: fix a hang that occurs when entering the delete prompt. problem appeared after the mythtv-pip merge.

[skamithi]

(In [15983]) Refs #2287. merges 15643:15965 from trunk to mythtv-vid

[skamithi]

(In [15994]) Refs #2287. mythtv-vid: minor cleanup after last update from trunk.

[skamithi]

(In [16222]) Refs #2287. merge 15965:16219 from trunk to mythtv-vid

[skamithi]

(In [16264]) Refs #2287. mythtv-vid: fix mythfrontend hang when exiting EPG

[skamithi]

(In [16402]) Refs #843. Refs #2287. mythtv-vid: part of [13617] was causing a hang when calling EPG or embedded playbackbox. modified the code to prevent hang. also fix bookmark setting issue when swapping PiP/PbP.

[skamithi]

(In [16466]) Refs #2287. mythtv-vid: prevent a MythPopupBox from displaying a mythbackend connection failure while activating the osd menu. it causes a hang.

[skamithi]

(In [16482]) Refs #2287. mythtv-vid: fix problem where inuseprograms table had stale entries after playback. prevented deletion of shows.

[skamithi]

(In [16520]) Refs #2287. merge 16219:16508 from trunk to mythtv-vid. contrib directory not merged properly. will resolve that soon.

[skamithi]

(In [16547]) Refs #2287. continuation of trunk -> mythtv-vid merge in [16520]. contrib dir merge successful now.

[skamithi]

(In [16548]) Refs #2287. merge -r 16508:16547 from trunk to mythtv-vid

[skamithi]

(In [16582]) Refs #2287. mythtv-vid. remove [16466]. breaks live tv playback. apply patch from shane shrybman that fixes channel changing problems.

[skamithi]

(In [16692]) Refs #2287. merge 16508:16686 from trunk to mythtv-vid

[skamithi]

(In [16727]) Refs #843 Refs #2287. mythtv-vid: adding some verbose messages regarding picture by picture and the pip state variable.also resolve some minor errors in the playercontext code

[skamithi]

(In [16781]) Refs #2287. merge 16686:16777 from trunk to mythtv-vid

[skamithi]

(In [17295]) Refs #2287. merge 16777:17292 from trunk to mythtv-vid

[skamithi]

(In [17296]) Refs #2287. fix compile issues after last merge from trunk.

[skamithi]

(In [17297]) Refs #2287. undo an verbose statement left after doing some troubleshooting.

[skamithi]

(In [17386]) Refs #843, Refs #2287. mythtv-vid: changes made to the preview video for recorded shows in mythtv-vid branch don't work after qt4 update . working on fixing this. for now disabling the feature.

[skamithi]

(In [17393]) Refs #843, Refs #2287. mythtv-vid: fixes preview video changes after qt4 upgrade.

[skamithi]

(In [17394]) Refs #2287, Refs #5324. mythtv-vid: commit patch(1) This is only used in one place (PiP) and will become increasingly tricky to maintain. Simply delete and recreate instead.

[skamithi]

(In [17395]) Refs #5324, Refs #2287, mythtv-vid: commit patch(2): Remove offscreen rendering code

[skamithi]

(In [17396]) Refs #5324, Refs #2287. mythtv-vid: commit patch (3) of #5324. Rename key variables in line with VideoOutput? to make opengl code easier to follow.

[skamithi]

(In [17397]) Refs #5324, Refs #2287. fix compilation issues when opengl video is disabled. incorporates patch(4) of #5324 by providing a better fix then [13916]

[skamithi]

(In [17414]) Refs #5324, Refs #2287. mythtv-vid: commit patch(5) and patch(13). Move OpenGL state management to OpenGLContext remove [15453] as patch(5) resolves this issue in a better way.

[skamithi]

(In [17419]) Refs #5324, Refs #2287. mythtv-vid: commit patch(7). opengl HD fix.

[skamithi]

(In [17430]) Refs #5324, Refs #2287. mythtv-vid: opengl thread safety locking improvements. patch(8)

[skamithi]

(In [17431]) Refs #5324, Refs #2287. mythtv-vid: simplify openglvideo::teardown, implement pixelbufferobject support. commits patch (9),(10),(15)

[skamithi]

(In [17432]) Refs #2287. mythtv-vid: make opengl vsync use some of the recent opengl changes commited. remain contextlocker convenience function to openglcontextlocker. remove extra processframe call in opengl code from videoout_xv.cpp that was introduced by mistake.

[skamithi]

(In [17441]) Refs #2287. merge 17292:17440 from trunk to mythtv-vid.

[skamithi]

(In [17510]) Refs #2287. merge 17440:17508 from trunk to mythtv-vid

[skamithi]

(In [17521]) Refs #5324 , Refs #2287, mythtv-vid: opengl renderer. Move the picture control code from OpenGLVideo to OpenGLContext so that the picture attributes can be shared between the main video, PiP and osd. (patch 11)

[skamithi]

(In [17523]) Refs #5234, Refs #2287, mythtv-vid: don't use glx1.2 calls when glx 1.3 is supported. patch(12).

[skamithi]

(In [17530]) Refs #2287. minor mythtv-vid fixes. uncomment some code and fix incorrect action when exiting tv browse mode.

[skamithi]

(In [17731]) Refs #2287. mythtv-vid: minor fix. fix typo in playercontext.cpp. causes onefield to be applied on all videos instead of only on sofware scaled video.

[skamithi]

(In [17760]) Refs #843. Refs #2287. mythtv-vid: make opengl PBP behave like the Xv PBP. OSD does not overlay over both pictures. OSD stays within only the active picture.

[skamithi]

(In [17765]) Refs #5324, Refs #2287. mythtv-vid. Tidy up some locking and ensure GL_PACK_ALIGNMENT is set to the correct value.

[skamithi]

(In [17766]) Refs #843, Refs #2287. mythtv-vid. fix swapping between PIP & PBP after applying [17760]

[skamithi]

(In [17775]) Refs #5324 , Refs #2287. mythtv-vid. patch(17) Tidy up YV12 packing code and add proper packing/upsampling of interlaced YV12 frames

[skamithi]

(In [17787]) Refs #843, Refs #2287. mythtv-vid. enable 2% vertical scaling for the hardware rendered preview player and software scaled PiP.

[skamithi]

(In [17848]) Refs #2287. 17508:17847 from trunk to mythtv-vid

[skamithi]

(In [17850]) Refs #2287. apply [17849] from trunk to mythtv-vid.

[Nigel]

(In [17895]) svn merge -r17851:17894 blah:.../trunk/mythtv . Have only tested compilation, not other possible breakage. Refs #2287

[skamithi]

(In [17904]) Refs #2287, Refs #843. fix some wrong logic introduced by me during a mythtv-vid merge [ 17848].

[skamithi]

(In [17905]) Refs #2287. mythtv-vid: Commit hack from mark kendall that allows opengl to work on the latest nvidia drivers.

[skamithi]

(In [17954]) Refs #2287. merge [17894]:[17952] from trunk to mythtv-vid

[skamithi]

(In [17985]) Refs #2287 merge 17952:17984 from trunk to mythtv-vid

[skamithi]

(In [17993]) Refs #2287, mythtv-vid: typo found after last merge.

[skamithi]

(In [17994]) Refs #843, Refs #2287. mythtv-vid: remove code I put in that was supposed to prevent resizing of opengl rendered picture during tv playback startup. It doesn't work. currently if you start EPG at tv playback you'll see nothing in the resized tv window and when you exit the EPG the picture is not resized again properly. Looking for a different solution that will work.

[skamithi]

(In [18013]) Refs #5324, Refs #2287. apply patch 19. Rewrite of deinterlacing code. Combine YUV2RGB and deinterlacing. Use 3 different fragment programs for progressive, top field and bottom field. Now properly deinterlaces luma and chroma channels and much faster. FrameBufferObjects? no longer a requirement. patch 18 and 18a, are already applied. this was done through [17954].

[skamithi]

(In [18050]) Refs #5324, Refs #2287. mythtv-vid: Add a GPU based YADIF deinterlacer .patch(20)

[skamithi]

(In [18456]) Refs #2287. merge 17984:18441 from trunk to mythtv-vid

[skamithi]

(In [18460]) Refs #2287. fix compile issue after last sync with trunk. fix EmbedInWidget failure on opengl renderer due to [18069].

[danielk]

(In [18899]) Refs #2287. Merge r18888:18898 from trunk to mythtv-vid, see also [18890].

[danielk]

(In [18911]) Refs #2287. Merges some misc changes from mythtv-vid to trunk.

[danielk]

(In [18916]) Refs #2287. Unremove copy_pixels_to_texture() function removed in [18911], its still being used in head when experimental opengl video is enabled.

[danielk]

(In [18919]) Refs #5324. Refs #2287. Applies Mark's mythtv-vid-22 patch, with some modifications.

[danielk]

(In [18920]) Refs #2287. Some audio refactoring that was done in mythtv-vid. This cleans up the volume/muting api a little bit.

[danielk]

(In [18958]) Refs #2287. Merges r18898:18957 from trunk to mythtv-vid

[danielk]

(In [18980]) Refs #2287. Refactor of NVP to remove duplication of pointers to things already contained in the PlayerContext. (mythtv-vid branch)

[danielk]

(In [18987]) Refs #2287. Begins cleaning up PlayerContext locking. There are still over a hundred unlocked uses, but the most common uses are now covered.

If your running with this branch right now, please run the frontend with '-v playback', this will print potentially useful lock debugging info in case you manage to trigger a deadlock.

[danielk]

(In [18992]) Refs #2287. Fix the pause lockups seen after the changes in [18990]. Note: There is still one unsafe section of code for the pause but it is well marked and isolated.

[skamithi]

(In [19030]) Refs #2287 Merges r18957:19029 from trunk to mythtv-vid

[danielk]

(In [19031]) Refs #2287. Revert portion of [18069] which broke several OSD renderers in mythtv-vid branch (my fault).

[danielk]

(In [19032]) Refs #2287. Refs #5324. Applies Mark Kendal's bicubic filter patch #22 to mythtv-vid.

[danielk]

(In [19033]) Refs #2287. Refs #5324. Applies Mark Kendal's disable textures patch #24 to mythtv-vid; this adds a DisableTextures method to OpenGLContext to match the EnableTextures method and uses it in openglvideo instead of raw OpenGL.

[danielk]

(In [19034]) Refs #2287. Refs #5324. Applies Mark Kendal's mmx packing patch (no.25), this lets mythtv use MMX more often when it's available for the OSD YV12 packing.

[danielk]

(In [19035]) Refs #2287. Refs #5324. Applies Mark Kendal's resize stage patch (no.26), this allows deinterlacing to work with resized opengl video frames.

[danielk]

(In [19036]) Refs #2287. Refs #5324. Correct field reversal that was making OpenGL bobdeint look bad. (patch 27 from Mark Kendal)

[danielk]

(In [19041]) Refs #2287. Fix main playback window aspect ratio problem in mythtv-vid. Aspect ratio adjustments were being applied twice...

[danielk]

(In [19042]) Refs #2287. Make sure playingInfo is locked when we dereference it.. the missing locks were causing some segfaults.

[danielk]

(In [19045]) Refs #2287. Fix a locking problem introduced by [19042], basically the commBreakMapLock is non-recursive so it was being locked twice by NVP::AutoCommercialSkip() after I added a missing locking in NVP::ClearAfterSeek() which was being called with and without a lock being held but calls SetCommBreakIter() which requires the lock to be held. Now ClearAfterSeek is only called when the lock isn't being held.

[danielk]

(In [19056]) Refs #2287. Fixes input name display bug in mythtv-vid by merging fix from trunk (probably missed in an earlier merge).

[danielk]

(In [19086]) Refs #2287. Merges r19029:19085 from trunk to mythtv-vid

[danielk]

(In [19087]) Refs #2287. Fixes another skipping deadlock in mythtv-vid.

[danielk]

(In [19089]) Refs #2287. A hacky fix for the PBP aspect ratio problem in mythtv-vid

[danielk]

(In [19095]) Refs #2287. Compile fix for when OpenGL Video is disabled

[danielk]

(In [19096]) Refs #2287. Parellelize OpenGLContext creation; this was causing a 130 ms slowdown in playback startup.

[danielk]

(In [19097]) Refs #2287. Shave another 90 ms off video playback startup time with a wait condition for NVP::playing

[danielk]

(In [19098]) Refs #2287. Fix a verbose statement

[danielk]

(In [19099]) Refs #2287. Various small VideoOutputXv cleanups.

[danielk]

(In [19118]) Refs #2287. Don't make the OpenGL window visible unless we're using it for display (as opposed to for V-Sync.)

[danielk]

(In [19120]) Refs #2287. Use QReadWriteLock instead of QMutex for player array locking.

[danielk]

(In [19121]) Refs #2287. Add context to UNMUTE message so it always goes to the right player.

[danielk]

(In [19129]) Refs #2287. Converts most of the event loop in the TV class to Qt timer events.

Five untriggered events remain, two need to be initiated by nvp actions, two I don't understand the code well enough, and one I just haven't looked at yet. I'm only checking these remaining untriggered events every 250 ms rather than every 50 ms; since all the stuff that needs to be snappy is now triggered immediately.

The code in tv_play.cpp still needs to be made pretty, more refactoring into PlayerContext is required, and this is probably buggier than yesterday's code, but it's on it's way to being better. The UI feels significantly snappier than it did a week ago.

[danielk]

(In [19172]) Refs #2287. Merges r19085:19164 from trunk to mythtv-vid

[danielk]

(In [19178]) Refs #2287. Merges r19164:19177 from trunk to mythtv-vid

[danielk]

(In [19182]) Refs #2287. Some VDPAU tweaks for the mythtv-vid branch.

Note there are still some problems

• Resolution changes are not handled by the VDPAU code
• VDPAU is treated as a subset of XVideo, videout_xv.cpp may be simpler if it is treated as a subtype outside of the XVideo range.
• All the other known problems, color in OSD, quality of MPEG-2 decoding, etc.

[danielk]

(In [19183]) Refs #2287. Cleanup ApproveDeintFilter() & include vdpau in base class version

[danielk]

(In [19184]) Refs #2287. Since both VDPAU & OpenGL renderers need the context reinitialized on resolution changes, special cases these in VideoOutputXv::InputChanged()

[danielk]

(In [19186]) Fixes #5950. Refs #2287. Backports VDPAU tweaks/fixes from mythtv-vid to trunk [19182],[19183],[19184].

[danielk]

(In [19187]) Refs #2287. Merges r19177:19186 from trunk to mythtv-vid.

[danielk]

(In [19188]) Refs #2287. Fix segfault in mythcommflag & mythtranscode in mythtv-vid

[danielk]

(In [19273]) Refs #2287. Merges r19186:19272 from trunk to mythtv-vid.

[danielk]

Updated task description

[danielk]

(In [19277]) Refs #2287. Don't try to restart LiveTV if we have hit an irrecoverable error in starting LiveTV.

Basically the reasoning is that if we hit an irrecoverable error once, we will likely hit it again and again on each retry.

This could easilly be extended to retry a few times before giving up, but I've only seen this occur with truly irrecoverable errors, such as the recording directory not being mounted. In trunk we set wantsToQuit to true for these irrecoverable errors, but that is not correct, since wantsToQuit indicates that the user wants to quit the app, and when there is an irrecoverable error we really want to show an error dialog then quit..

AFAIK: This problem really only occurs in practice because we haven't ported the error dialogs in the NVP and TV classes to Qt4 yet, but even once we do -- this change will prevent infinite loops if we forget to implement an error exit dialog for some irrecoverable eror.

[danielk]

(In [19288]) Refs #2287. Make the position map transfer code for in progress recordings a bit more efficient with a GetKeyframePositions() call in the recorder, and make GetKeyframePosition() work as documented (returning closest prior position if the frame is not found in the map) instead of returning -1 for non-keyframes.

[danielk]

(In [19290]) Refs #2287. Some cleanup of tv_play.cpp based on a diff with trunk.

Other than a invalid negation that snuck into REC_PROGRAM checking for pseudoLiveTV, this just removes some redundant code in TV::customEvent()

[danielk]

(In [19296]) Refs #2287. Moves pseudoLiveTV and playback speed change event checking out of the generic timer event in TV clas (mythtv-vid).

[danielk]

(In [19298]) Refs #2287. Eliminates the TV class' generic timer used for pieces of code that used to be in event loop but didn't have any event mechanism of their own yet.

[danielk]

(In [19299]) Refs #2287. Return PlayerReadLock in TV::HandleEndOfPlaybackTimerEvent, broken by [19298].

[danielk]

(In [19301]) Refs #2287. A few minor fixes.

[danielk]

(In [19311]) Refs #2287. Partially revert [19301], we still need to run processEvents() regularly in TV::StartTV() even if there are no state changes. This is needed for keyboard and other event processing on at least some Qt 4.x releases. Possibly a better solution is to enter the event loop for the thread running StartTV(), but that requires a bit more refactoring...

[danielk]

(In [19312]) Refs #2287. Revert extra files commited in [19311], these are from something else that wasn't intended for commit yet.

[markk]

(In [19313]) Refs #2287. Disallow doubleprocess deinterlacers as fallbacks (mythtv-vid)

[markk]

(In [19314]) Refs #2287. Move some more OpenGL state management into OpenGLContext - removes some redundant OpenGL state calls. Tidy up some comments. (mythtv-vid)

[markk]

(In [19316]) Refs #2287. Initial cross-platform OpenGL support in util-opengl (mythtv-vid).

Primarily adding entry points and defines for any functionality newer than OpenGL 1.1 (unavailable on windows). Also add hooks for required windows and Mac OS X extensions and swap_control on Linux and windows (available by default on OS X).

[danielk]

(In [19318]) Refs #2287. Clean up autoskip settings code in mythtv-vid.

[markk]

(In [19321]) Refs #2287. Add cross-platform support to OpenGLContext (mythtv-vid).

Push all window system specific functionality into sub-classes - OpenGLContextGLX (Linux/X11), OpenGLContextWGL (windows xp/vista) and OpenGLContextAGL (Mac OS X).

[danielk]

(In [19327]) Refs #2287. Use our X11 lock when processing Qt's X11 calls.

This is theoretically a problem in trunk too, Qt can make an X11 call at the same time as VideoOutputXv. But it becomes a real segfault producer in mythtv-vid because we have much fewer usleep() calls scattered throughout the code.

[danielk]

(In [19328]) Refs #2287. Add a mutex for m_positionMap.. I saw SEGFAULTs on this in testing.

[danielk]

(In [19329]) Refs #2287. Add some useful debugging to GetFramesWritten() and make sure we return -1 if the backend returns and error.

[danielk]

(In [19330]) Refs #2287. Make keyframe map syncing a bit more efficient. First we allow the NVP to pass -1 as the end keyframe to request, which means we want all keyframes greater than the start keyframe this avoids 1 of the two network roundtrips. Second, we use QMap::lowerBound() to find the first keyframe greater than or equal to the start keyframe, and then iterate until we reach the end of the map, this doesn't change the big O performance but makes the code about a 15 times faster in practice.

[danielk]

(In [19331]) Refs #2287. Improve NVP error recovery a bit & improve the debugging macros.

[danielk]

(In [19332]) Refs #2287. Fix a few problems with PIP in mythtv-vid

[danielk]

(In [19334]) Refs #2287. Restore XVideo playback when OpenGL stuff is compiled in and reapply an nvidia driver bug workaround. These was accidentally broken/reverted by [19321].

[danielk]

(In [19335]) Refs #2287. Merges r19273:19334 from trunk to mythtv-vid

[danielk]

(In [19336]) Refs #2287. Make sure we have the video filters lock before applying a video filter on a frame for the PIP and before deleting the NVP. Otherwise we get segfaults on program changes in the PIP.

[danielk]

(In [19345]) Refs #2287. Cleans up PBP startup.

Use JumpToFrame() to restart NVPs in the same place in the file, as we do for PIP. Use the currently active NVP's OSD in PBP mode, unless the OSD for the main buffer is a full screen OSD (None are at the moment, so that part is just a stub in playercontext.cpp).

[danielk]

Adds Multiple PiP window support.

[danielk]

The multi-pip-v1.patch has a lot of extraneous changes including debugging that needs to be ripped out, NVP/TV error handling is broken, player teardown is not done correctly for PIP's, and this likely won't compile with video output classes other than XVideo. But, it does finally allow for multiple Picture-in-Picture windows...

Note: After I clean this up I plan to extend PBP as well, so that you can do things like look at all the channels on the current multiplex at once. After that I plan to allow audio output follow you to the currently active player...

[danielk]

(In [19367]) Refs #2287. Fixes buffer overflow in OSDListBtnType::InitItem().

[danielk]

(In [19373]) Refs #2287. Merges r19334:19372 from trunk to mythtv-vid

[danielk]

updated patch

[danielk]

(In [19375]) Refs #2287. VideoOutput portion of multi PIP patch.

[danielk]

updated patch

[danielk]

(In [19376]) Refs #2287. Fixes two memory leaks. (Note the jump to program code produces segfaults, but the patch to fix that will be applied seperately.)

[danielk]

(In [19377]) Refs #2287. Fixes Jump-to-program segfault.

[danielk]

(In [19378]) Refs #2287. Applies multi-pip patch, sans debugging stuff.

[danielk]

Debugging portion of multi-pip patch.. if you want to fix any of the problems, this is a good thing to apply first..

[danielk]

(In [19388]) Refs #2287. Reset null video vars on PIPTeardown so PIP swaps don't fail due to improper settings.

[danielk]

(In [19389]) Refs #2287. Removes uneccessary function call which was causing deadlocks on PIP window channel changes.

[danielk]

(In [19391]) Refs #2287. When we enter TV::DoEditSchedule() with a PIP or second PBP window as the active window, use that window for any playback in grid functionality. Also if the main window is the active window, pause any PIP/PBP windows and don't render them until we return to normal playback.

[danielk]

(In [19393]) Refs #2287. Allow TV::StartTV() to exit before the worst case startup timer expires if we actually see playback start. This is an optimization for the case where you accidentally start the wrong recording and want to exit as soon as you realize you've made a mistake...

[danielk]

(In [19394]) Refs #2287. Improves debugging output for TV::SetActive()

[danielk]

(In [19395]) Refs #2287. Couple small changes, these don't change functionality, but one silences confusing debugging statements, the other speeds up state transition handling.

[danielk]

(In [19401]) Refs #2287. Special case OSD margins and font size in PBP mode.

[danielk]

(In [19402]) Refs #2287. Fix exit code in end-of-file detection.

[danielk]

(In [19403]) Refs #2287. Sanitize nvp locking in TV class. (For now this is just a mutex, probably will be RW lock later once I'm sure there are no deadlocks hiding in there.)

[danielk]

(In [19404]) Refs #2287. Fixes some segfaults in the OpenGL video rendering.

[danielk]

(In [19405]) Refs #2287. Refactors PxPSwap().

[danielk]

(In [19408]) Refs #2287. Cleanup of PlayerContext (with resulting changes in other files from method renaming.)

[danielk]

(In [19413]) Refs #2287. Since a Picture-by-Picture window can have it's own OSD, move the dialogname from the TV class to the OSD class.

[danielk]

(In [19414]) Refs #2287. Merges r19372:19413 from trunk to mythtv-vid

[danielk]

(In [19417]) Refs #2287. Merges mythtv-vid to trunk.

There are a number of changes, probably the most user visible ones are PiP related, but there is also groundwork laid for OpenGL video rendering on OSX and MS Windows, for aspect ratio & zoom on OSX, and a number of operations are more responsive.

Note: Even though this changeset has received a fair bit of testing, it is a 1.1MB patch and there are bound to be things we overlooked. Please report any compilation problems to the mythtv-dev mailing list and open tickets with a repro description, theme used and backtrace for any segfaults or deadlocks.

[danielk]

(In [19418]) Refs #2287. Merges trunk to mythtv-vid

[danielk]

(In [19419]) Refs #2287. Quick compile fix for vdpau

[danielk]

(In [19459]) Refs #2287. Fixes a locking problem using a backtrace & problem description reported by Mark Kendall.

[danielk]

(In [19500]) Refs #2287. Remove a debugging value that snuck into svn.

[markk]

(In [19513]) Refs #2287. Merges r19419:19512 from trunk to mythtv-vid.

[markk]

(In [19535]) Refs #2287. Add a new cross-platform OpenGL video renderer that provides consistent playback across Linux, Windows XP, Vista and Mac OS X.

There are some small remaining issues with xrandr on Linux and PBP sizing across the board.

N.B. Due to a lack of altivec YV12 packing code, performance may be poor on non-intel Macs. N.B. I've been unable to test a windows build for the last few weeks.

[markk]

(In [19544]) Refs #2287. Remove code duplication between XVideo and OpenGL video renderers.

[markk]

(In [19546]) Refs #2287. Fix Mac OS X compilation with patch from Stuart Auchterlonie.

[markk]

(In [19548]) Refs #2287. Merges r19514:19546 from trunk to mythtv-vid.

[markk]

(In [19549]) Refs #2287. Remove more code duplication between XVideo and OpenGL.

[markk]

(In [19664]) Refs #2287. Merges r19547:19663 from trunk to mythtv-vid.

[markk]

(In [19675]) Refs #2287. Fix OpenGL video preview in epg. (mythtv-vid)

[markk]

(In [19677]) Refs #2287. Enable color controls for non XVideo builds. (mythtv-vid)

[markk]

(In [19847]) Refs #2287. Merge r19664:19846 from trunk to mythtv-vid.

[markk]

(In [19889]) Refs #2287. Small fixes and refactoring for OpenGL rendering code.

[markk]

(In [19890]) Refs #2287. Fix OpenGL bobdeint in mythtv-vid.

[markk]

(In [19891]) Refs #2287. Re-enable cpu based bob deinterlacer after move to cross-platform OpenGL video renderer. mythtv-vid

[markk]

(In [19924]) Refs #2287 (mythtv-vid). Add earlier checks for OpenGL extension suppport when adding OpenGL filters. Aimed mostly at failing earlier and faster on Intel chipsets.

[markk]

(In [19972]) Refs #2287. Merge r19847:19971 from trunk to mythtv-vid.

[markk]

(In [19973]) Refs #2287. Respect Xinerama window sizing in VideoOutputOpenGL (consistent with VideoOutputXVideo).

[markk]

(In [19982]) Refs #2287, Refs #5257. Fix OpenGL deinterlacing of bottom field first streams.

[markk]

(In [20077]) Refs #2287. Merge -r19972:20076 from trunk to mythtv-vid.

[markk]

(In [20081]) Refs #2287. Small OpenGL texture filtering fix.

[danielk]

Any remaining issues should get their own tickets.