Original commit message from CVS:
2005-07-07 Andy Wingo <wingo@pobox.com>
* sys/v4l/v4lsrc_calls.h:
* sys/v4l/v4lsrc_calls.c: Remove sync-related stuff.
(gst_v4lsrc_get_fps_list): Moved here from gstv4lsrc.c.
(gst_v4lsrc_buffer_new): Totally derive from GstBuffer.
* sys/v4l/v4l_calls.h: Cast to V4lElement.
* sys/v4l/v4l_calls.c: Header loc fixen, don't load mjpeg, all
v4lelements are sources.
* sys/v4l/gstv4lxoverlay.h:
* sys/v4l/gstv4lxoverlay.c:
* sys/v4l/gstv4ltuner.h:
* sys/v4l/gstv4ltuner.c: Header loc fixen.
* sys/v4l/gstv4lsrc.h:
* sys/v4l/gstv4lsrc.c: Crucial GPL update. Clean up a bit, port to
PushSrc/BaseSrc. Removed most sync-related properties, videorate
or something should handle that. Made a live source.
* sys/v4l/gstv4lelement.h:
* sys/v4l/gstv4lelement.c: Derive from GstPushSrc. No more
signals. Some cleanups.
* sys/v4l/gstv4lcolorbalance.h: Interface header update.
* sys/v4l/gstv4l.c: Don't register v4lelement, or the jpeg/mjpeg
stuff.
* sys/v4l/Makefile.am: Build everything except the jpeg/mjpeg
stuff.
* sys/Makefile.am (SUBDIRS): Hit the V4L crack pipe.
Original commit message from CVS:
* ext/vorbis/vorbisdec.c: (vorbis_dec_chain):
Debug.
* sys/v4l/v4lsrc_calls.c: (gst_v4lsrc_grab_frame):
If we got a state change in the _get handler, don't return success.
Original commit message from CVS:
* sys/v4l/v4lsrc_calls.c: (gst_v4lsrc_try_capture):
Don't actually error out if we get another return value than
-EINVAL. Opposite to what I first thought, drivers have random
return values for this, although -EINVAL is the expected return
value. Since this is not fatal, we shouldn't use
GST_ELEMENT_ERROR() but just GST_ERROR_OBJECT().
Original commit message from CVS:
compatibility fix for new GST_DEBUG stuff.
Includes fixes for missing includes for config.h and unistd.h
I only ensured for plugins I can build that they work, so if some of them are still broken, you gotta fix them yourselves unfortunately.
Original commit message from CVS:
Some final fixes for the v4lsrc elements.
* remove software sync thread (use GST_ELEMENT_THREAD_SUGGESTED instead)
* make all src elements threadsafe
* fix num_buffer argument setting in v4l2src (VIDIOC_S_PARM)
* re-add bufsize (RO) for v4lmjpegsrc
* fix the A/V sync calculation in all elements (spvf=GST_SECOND/fps, not GST_SECOND*fps)
* probably some more crap....
With all this, it actually works quite well. The TODO files describes the
next steps in order to make a full-featured video recorder based on these
elements and GStreamer (bottom). Making a simple recorder should be fairly
easy now, btw.
Original commit message from CVS:
make v4l element do all calls in one thread, this makes performance much better (because of kernel locking, more threads doesn't make sense anyway)
Original commit message from CVS:
This patch fixes some issues caused by design issues in video4linux, adds
some nicety to video4linux2 plugins and does some more evil stuff:
* video4linux doesn't tell us which formats are supported by a card, so
the only way to know this is by simply trying it out. This patch adds that.
* v4lmjpegsink didnt have a bufferpool yet - is integrated now.
* all copy() bufferpool functions have been removed since they're not needed.
* v4lmjpegsink doesnt have a free() function, because hen playing the frames,
all this is already handled. When the frame is not played, nothing has to
be done. In total, the function is not needed.
* adds a get_caps() function to v4l2src
* some minor crap
Original commit message from CVS:
This patch:
1) would provide v4l2 plugins, if only the CVS server wouldn't bork on all commits :-(. This patch will come later on in a separate commit when SF fixes her repository.
2) it fixes capsnego for all the video4linux1 plugins
3) it rewrites the debugging of all v4l1 plugins, which now make use of the standard gstreamer debugging features (gst-mask=...). This should make debugging video4linux1/gstreamer problems much easier then it used to be.
VS: ----------------------------------------------------------------------
Original commit message from CVS:
* removal of //-style comments
* don't link plugins to core libs -- the versioning is done internally to the plugins with the plugin_info struct,
and symbol resolution is lazy, so we can always know if a plugin can be loaded by the plugin_info data. in theory.
Original commit message from CVS:
Added pthread software sync for improved timestamps and pthread queue and sync control to make sure that frames are queued if we sync on them
