* We can actually tell libavformat to discard packets for streams that
we are not interested in. Found this in the ffplay code. This should hopefully
avoid the efficiency impact of using one AVFormatContext per stream.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31458 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Refactored NegotiateOutputFormat() and Decode() into two separate private
methods each, one for video and one for audio.
* Keep reading chunks when video decoding, until we have got a picture. This
gets us scrambled video instead of a black picture for h264 in mpegts.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31457 a95241bf-73f2-0310-859d-f6bbb57e9c96
context has AVPrograms. I gather this feature is for container streams
that contain multiple "groups of streams" like how it would work for
satilite mpeg-ts streams with multiple TV channels in one stream. For this
to be properly supported, we should extend the BMediaFile/Track API. For
now, the AVFormatReader uses the first program, if one is there. This
was also needed to get make mpegts demuxer work, but it is not yet enabled
for other reasons.
* Read more probe data. 1024 bytes were not enough to detect "mpegts" properly
for example.
* For now, I disabled the locking in the AVFormatReader hooks themselves,
this should not be necessary, though I hope libavformat is reentrant as
long as you have your own AVFormatContext for each thread. So far everything
hints that it is the case.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31454 a95241bf-73f2-0310-859d-f6bbb57e9c96
just architecture specific flags, call them just HAIKU_FFMPEG_DEFINES.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31432 a95241bf-73f2-0310-859d-f6bbb57e9c96
how everything currently worked. (Packet peaking was only done once in Init().)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31423 a95241bf-73f2-0310-859d-f6bbb57e9c96
container formats we support. I hope I have turned on only those that don't
have an implementation already (did we support ASF already?). I mostly tested
with AVI and that works reasonably well, but I only tested raw audio so far.
However, the backend for Marcus' avi_reader is much nicer than FFmpeg, so I
think it should stay and have disabled the support for AVI in AVFormatReader.
A big disappointment is that MPG containers only give scrambled video. It may
be a problem of the AVCodecDecoder, but I am not so sure. After all, it works
fine for other container formats. If I am not mistaken, VLC also does not use
the mpeg demuxer from FFmpeg. :-\ On top of that, libavformat detects the
time_base and stream duration wrongly for one of my test MPGs.
As for the implementation: Although seeking in libavformat happens for an
individual stream, in reality, the packets for all other streams need to be
flushed (that's also what happens in the libavformat tutorials I've seen).
Since our MediaKit API allows to seek tracks indivually, this is of course
a no-go, since then all other tracks would be out of sync. My solution is to
simply open the demuxer once for each stream and then completely ignore the
packets of the respective other streams. This also works around the problem
that libavformat is unable to provide packets by stream, requiring the
API user to queue the packets that he needs not now, but later for other
streams. It also means we have to no memory copies, since we can directly
use the packet buffer until the next call to GetNextChunk().
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31421 a95241bf-73f2-0310-859d-f6bbb57e9c96
Re-use the kernel emulation of the FS shell to implement a FUSE module
for BFS. In theory, it could use any Haiku file system implementation
just like the FS shell itself, but currently only the BFS code supports
being compiled as an FS shell module. The build target is <build>bfs_fuse
On Ubuntu, you would do this:
sudo apt-get install libfuse-dev
jam \<build\>bfs_fuse
mkdir path/to/moundPoint
path/to/bfs_fuse /dev/sdaX path/to/mountPoint
And that will mount your BFS volume at /dev/sdaX
read/write. As this is new code, you should be careful with your
data and expect potential data loss. Not that I don't trust Raghu's
code, but a warning is always better with such things. So far, my
tests have been successful and it is great to finally have write
support for BFS in Linux! Thanks a lot, Raghu, for your great work!
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31409 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Moved stuff from testing in Sniff() into class members.
* Added function to gfx_utils() that converts an FFmpeg pix_fmt to color_space.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31402 a95241bf-73f2-0310-859d-f6bbb57e9c96
* The URLProtocol idea seems not to work out, so I removed that code, but
the other idea of setting up a ByteIOContext actually works, once I seek
back to the beginning of the stream after reading the initial probe buffer,
we may also offset the buffer pointer in the ByteIOContext to where we
have already read, but I am not sure of possible side effects of that.
We can now probe for the correct demuxer and detect streams.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31379 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Turn off tracing in the AVCodecDecoder which I accidentally turned on in
a previous commit.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31366 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Removed inclusion of internal.h from libavcodec.h, it's not there in the
plain FFmpeg 0.5 version of the file.
The last item fixes the GCC2 build, at least AFAICT.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31364 a95241bf-73f2-0310-859d-f6bbb57e9c96
unified the Reader and Decoder plugins and renamed the add-on to "ffmpeg".
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31358 a95241bf-73f2-0310-859d-f6bbb57e9c96
"ffmpeg" and export multiple plugin types from the same plugin (Decoder and
Reader).
* Work in progress on an libavformat based Reader plugin, does nothing yet.
* Moved config.h from libavcodec subfolder up one level, so that it's used
by libavformat as well. Adjusted libavutil/common.h accordingly.
* Turned off GPL code in config.h
* Turned off BeOS muxers in config.h
* Turned on HAVE_THREADS and HAVE_PTHREADS, although that is nowhere used
in the ffmpeg code (it appears).
* Indentation cleanup in avcodecplugin.h
I have built this with GCC4, but last night I built libavformat.a with GCC2
so I am hoping this doesn't break the GCC2 built.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31355 a95241bf-73f2-0310-859d-f6bbb57e9c96
setting and made them available in the context menu again. During my testing,
I couldn't find any weird behavior, only the window dragging should still be
a problem on slower computers when multiple mouse messages piled up.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31324 a95241bf-73f2-0310-859d-f6bbb57e9c96
write 0 bytes after doing a successful copy of an attribute. Since
fs_write_attr() was actually ignoring the position argument, this would just
clobber attributes and truncate them back to 0 bytes. This was fixed in the
previous commit, however, it should be noted that if the buffer which
copy_attributes() uses were too small, writing attributes which live in the
"small data section" iteratively would not work because of a current BFS
limitation.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31310 a95241bf-73f2-0310-859d-f6bbb57e9c96
* As even the author of fs_attr.c was trying to use fs_write_attr() in an
unsupported way in the cp copy_attributes() implementation, try to be
more forgiving and support writing attributes at an offset. The method is
required to behave inconsistent in that replacing bytes at offset 0 is
not supported as replacing bytes at any other offset. Writing at offset 0
will clobber the existing attribute to stay compatible with BeOS.
NOTE: BFS itself still does not support writing attributes at an offset which
are withing the "small data section". To work around this problem, programs
which copy attributes in a loop must make sure that their buffer is large
enough that such "small data section" attributes require only one loop
iteration.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31309 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Added "Force Aspect Ratio" to Video menu and context menu.
* Removed some code duplication in handling the video size messages.
* Added track selection sub-menus to the context menu, replacing the IMHO
not so important interface settings items.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31302 a95241bf-73f2-0310-859d-f6bbb57e9c96
Totally "gapless", unlike any other media player I've ever used.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31301 a95241bf-73f2-0310-859d-f6bbb57e9c96
asking what to do with the debugged team. Killing it is not yet implemented.
Then it notifies the application about the quit request. That in turn tears
down the TeamDebugger and if no other ones are left, it quits itself.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31201 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Use insets around everything. The visual change are insets also around the
source code view. Another option would be to put the source code view on the
right/bottom, attaching the window border and using a B_DOCUMENT_WINDOW
resize knob on the window border. The register view could also use the
additional height of the buttons. However, I agree it's somehow more logical
to have the register view on the right side. Maybe only because of
convention... so it's possibly still up for debate.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31198 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Give access to BSplitLayout::SetInsets() and GetInsets() from BSplitView.
* Give access to BSplitView::SetInsets() from BSplitLayoutBuilder and
LayoutBuilder.
* Some automatic white-space cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31197 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Implemented a CPU feature detection function in AppServer.cpp.
The results are put into the global variable gAppServerSIMDFlags.
* Implemented an SIMD accelerated version of the bilinear bitmap
scaling code that is the backend of BView::DrawBitmap(...,
uint32 options) used by the MediaPlayer to smoothly upscale
movies when no video overlay is available. The speed up is very
noticable and a Core 2 Duo @ 1.8 GHz can play at 1920x1200 now
without breaking a sweat. There is currently one SIMD version
implemented which uses MMX and plain SSE.
Very cool! Thanks a lot!
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31165 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Added _MinTabeSize() to BTabView. It makes sure that the BTabView displays
at least two tabs. In the future it should also add the room to display some
buttons for cycling the currently displayed tabs left/right if there is more
than can fit.
* In BTabView::Min/Max/PreferredSize(), use _MinTabSize() to compute the
respective size.
* Improve Tab rendering code so there are no ugly overlaps at the right edge
if the tabs falls directly on the view edge.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31157 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Add some spacing around most views in the top level layout.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31152 a95241bf-73f2-0310-859d-f6bbb57e9c96
in BWindow::_FindView(int32 token). This fixes the BColumnListViews in
DebugAnalyzer and Debugger to draw their frames properly.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31151 a95241bf-73f2-0310-859d-f6bbb57e9c96
back to the system default cursor. For that cursor token, the global cursor
manager returns NULL.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31138 a95241bf-73f2-0310-859d-f6bbb57e9c96
* BSplitView uses ControlLook to draw itself.
* Small refactoring in SplitLayout to implement IsAboveSplitter(BPoint).
* BSplitView updates the view cursor when the mouse is above a splitter.
* Standard splitter size is now 6 pixels, which is a bit easer to hit.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31137 a95241bf-73f2-0310-859d-f6bbb57e9c96
window of the view into the application thread. This solves the
race condition with asynchronous SetViewCursor and deleting the
cursor immediately afterwards for real.
* The ServerApp now requires a reference to the current cursor,
just in case...
* Added TODOs for caching the BView token, it's currently resolved
for every single BView call that talks to the server... not good!
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31133 a95241bf-73f2-0310-859d-f6bbb57e9c96
reference counting is maintained correctly in the app_server. While reviewing
this code, I have my doubts that my previous solution for handling pending
SetViewCursor() calls is always working as it is intended.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31131 a95241bf-73f2-0310-859d-f6bbb57e9c96
not be NULL. If the intention is to allow this, then there is a bug somewhere
else. Fixes booting on Macs, thanks to Zenja for testing this!
* Fixed what I believe to be an off-by-one error in BlockCount(). This also
makes it symetrical with SetBlockCount().
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31069 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Introduced a lock in GetDefaultDevice() and ReleaseDevice() as a quick
solution to the race condition in those functions. It could also use
proper atomic ref counting. Untested.
* Automatic white space cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31060 a95241bf-73f2-0310-859d-f6bbb57e9c96
'system' folder will be a clean copy of the source volume. Other
folders will be merged (as before), but in case a folder is in the
way of a link or file from the source volume, it will now be purged.
* Clarified the alert for non-empty target volumes, so it is very clear
what happens. (Maybe there ought to the be option to only copy the system
folder, though.)
* Fixed a problem with copying attributes in certain cases.
* Fixed the main GUI not resetting state properly after encountering an
error during the copy process.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@31042 a95241bf-73f2-0310-859d-f6bbb57e9c96
needed. I've added MediaPlugin* fields to Reader and Decoder plugin classes
which are set when the PluginManager hands out new instances. This way the
manager knows what plugin created the Decoder or Reader instance in the
Destroy*() methods and can decrease the reference count accordingly. Also added
some FBC stuffing to Decoder and Reader. All media plugins need to be recompiled,
in case anyone has some outside the Haiku tree.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30984 a95241bf-73f2-0310-859d-f6bbb57e9c96
necessary when compiling the code for example with -O1 or with the upcomming
GCC 4.4. Thanks a lot!
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30979 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Fixed remapping the Control keys.
* Adjusted the caps-lock shifted keys, in case someone decides to exchange
the left backspace key for the caps-lock key again.
Thanks a lot!
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30968 a95241bf-73f2-0310-859d-f6bbb57e9c96
Sorry about that! This is now like in the original patch and fixes the ticket
for real.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30926 a95241bf-73f2-0310-859d-f6bbb57e9c96
with the mouse, so that switching between using the keyboard and mouse is more
fluent. Fixes ticket #3971.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30918 a95241bf-73f2-0310-859d-f6bbb57e9c96
depends on CDRecord for mkisofs, which is used for building the
*-cd targets. Thanks!
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30917 a95241bf-73f2-0310-859d-f6bbb57e9c96
probably still be improved.
* Renamed EntryRefPlaylistItem to just FilePlaylistItem.
* Moved the "move into Trash" and "restore from Trash" implementation into
FilePlaylistItem. Also added what's needed to allow Tracker to restore the
entry itself.
* Refactored everything to make Playlist use PlaylistItems instead of
entry_refs and all that entails...
* The transition to virtualize PlaylistItems is not complete yet in the
Controller, since it still uses BMediaFile there. But it's much easier to
change that now.
* Objects which keep a PlaylistItem around do correct reference counting, but
some commands could be simplified if they were using references as well. It
still should work correctly, though, if I didn't miss anything. It should also
fix theoretical situations of encountering out-of-memory while messing with
the Playlist leading to inconsistent state between Undo/Redo and then leaking
items because of that.
* Added the discussed change that MediaPlayer checks it's own supported types
before rejecting a file by super type. (untested)
* When importing playlist items, the currently playling item is better
maintained during Undo/Redo.
* Some debugging code added in MediaTrackAudioSupplier, no functional changes.
* Fixed a number of coding style issues and automatic whitespace cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30834 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Disable the screen saver when playing in full screen mode.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30797 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Calling BTextView::SetText() will invoke Window()->UpdateIfNeeded(), maybe
since some recent changes only - we should perhaps look into this... (zooey),
but the problem was that we called BControl::SetValueNoUpdate() after that,
and thus we were drawing the color mark at the old offset and the invalidation
in SetValue() was rendered effectless. Fixes#3719.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30784 a95241bf-73f2-0310-859d-f6bbb57e9c96
index and not work on the current selection only.
* Reenable the code in MainWin that implements the shortcuts to trigger this
feature during playback. Now it uses the Undo/Redo stack and profits from
existing polish.
* Add missing fPlaylist locking in MainWin at several places. Should not have
caused any realworld problems, though.
* Add Undo/Redo shortcuts to MainWin and forward to Playlist window.
* Make sure the Playlist window opens on the same workspace as the player
window. This could be seen when launching MediaPlayer on one workspace, later
moving the window and then opening the Playlist window, it would open on
the original workspace. The new behavior additionally pulls the playlist
when it's already open and you invoke the "Playlist" menu item.
* Added the beginnings of a refactoring to make the Playlist and everything
else use a PlaylistItem class, instead of entry_refs directly. This stuff is
not yet used, though, just compiles.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30776 a95241bf-73f2-0310-859d-f6bbb57e9c96
implemented. It didn't reuse existing code and didn't integrate well. No
Undo/Redo except via Tracker, but not in the Playlist... some bugs as well
(Remove had same shortcut as Randomize, Didn't maintain current playback item
if last entry was removed)
* I need to reenable the main window short cut though. This is only temporary.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@30768 a95241bf-73f2-0310-859d-f6bbb57e9c96
Stephan Aßmus