Uncompressed RGB frames can be (usually are) bottom-up
layout in DirectShow, and the code to flip them wasn't
properly ported from 0.10. Fix it.
Fix post-processing of RGB buffers. We need a writable
buffer, but the requests pool is holding an extra ref.
This could use more fixing to use a buffer pool
The pseudo buffer pool code was using gst_buffer_is_writable()
alone to try and figure-out if cached buffer could be reused.
It needs to check for memory writability too. Also check map
result and fix map flags.
https://bugzilla.gnome.org/show_bug.cgi?id=734264
GetOverlappedResult() might never return with some drivers. Time out
after 1000 ms. We cannot really fix this without either:
1) Controlling the streaming thread so we can do CancelIo() from that
thread.
2) Switch to using IO completion ports.
Turns out that the reference implementation does this, hence we need to
mirror this behaviour. This typically happens with hardware that takes
some time to initialize.
Most important part here is special-casing "device busy" so the application
is able to provide better feedback when another application is using the
device.
* Make the driver write directly into each GstBuffer to avoid memcpy().
* Don't memset() the buffer before reusing it.
* Recycle memory by keeping two spare buffers. Two because the sink
downstream may keep a ref to the previous buffer.
Note that we align buffers on highest possible byte boundary (4096) so we
don't have to take into account what kind of alignment the driver requires.
Original commit message from CVS:
* sys/winks/gstksvideodevice.c (GST_DEBUG_IS_ENABLED, last_timestamp,
gst_ks_video_device_prepare_buffers, gst_ks_video_device_create_pin,
gst_ks_video_device_set_state, gst_ks_video_device_request_frame,
gst_ks_video_device_read_frame):
Guard against capturing old frames by keeping track of the last
timestamp and also zero-fill the buffers before each capture.
Only assign a master clock if the pin hasn't already got one.
Actually free buffers on the way down to avoid a huge memory leak,
as this was previously done when changing state to ACQUIRE downwards
and we now skip that state on the way down.
Add some debug.
* sys/winks/gstksvideosrc.c (DEFAULT_DEVICE_PATH, DEFAULT_DEVICE_NAME,
DEFAULT_DEVICE_INDEX, KS_WORKER_LOCK, KS_WORKER_UNLOCK,
KS_WORKER_WAIT, KS_WORKER_NOTIFY, KS_WORKER_WAIT_FOR_RESULT,
KS_WORKER_NOTIFY_RESULT, KS_WORKER_STATE_STARTING,
KS_WORKER_STATE_READY, KS_WORKER_STATE_STOPPING,
KS_WORKER_STATE_ERROR, KsWorkerState, device_path, device_name,
device_index, running, worker_thread, worker_lock,
worker_notify_cond, worker_result_cond, worker_state,
worker_pending_caps, worker_setcaps_result, worker_pending_run,
worker_run_result, gst_ks_video_src_reset,
gst_ks_video_src_apply_driver_quirks, gst_ks_video_src_open_device,
gst_ks_video_src_close_device, gst_ks_video_src_worker_func,
gst_ks_video_src_start_worker, gst_ks_video_src_stop_worker,
gst_ks_video_src_change_state, gst_ks_video_src_set_clock,
gst_ks_video_src_set_caps, gst_ks_video_src_timestamp_buffer,
gst_ks_video_src_create):
Remove ENABLE_CLOCK_DEBUG define, it's GST_LEVEL_DEBUG after all.
Get rid of PROP_ENSLAVE_KSCLOCK and always slave the ks clock to the
GStreamer clock, it doesn't seem to hurt and matches DirectShow's
behavior. As an added bonus we usually get PresentationTime set for
each frame, so we can expand on this later for smarter latency
reporting (by looking at the diff between the timestamp from the
driver and the time according to the GStreamer clock).
Use an internal worker thread for opening the device, setting caps,
changing its state and closing it. This way we're a lot more
compatible with drivers that rely on hacks to do video-effects
between the low-level NT API and the application. Ick.
Start the ks clock and set the pin to KSSTATE_RUN on the first
create() so that we'll hopefully get hold of the GStreamer clock
from the very beginning. This way there's no chance that the
timestamps will make a sudden jump in the beginning of the stream
when we're running with a clock.
* sys/winks/kshelpers.c (CHECK_OPTIONS_FLAG,
ks_options_flags_to_string):
Reorder the flags to match the headerfile order, and make the string
a bit more compact.
* sys/winks/ksvideohelpers.c (ks_video_probe_filter_for_caps):
Avoid leaking KSPROPERTY_PIN_DATARANGES.
Original commit message from CVS:
* sys/winks/gstksvideodevice.c (gst_ks_video_device_class_init,
gst_ks_video_device_set_state):
Don't set the pin state to KSSTATE_RUN from the streaming thread.
Skip KSSTATE_ACQUIRE when changing pin state downwards.
Be nice and specify G_PARAM_STATIC_STRINGS.
Remove unused finalize method.
* sys/winks/gstksvideosrc.c (DEFAULT_ENABLE_QUIRKS, PROP_ENABLE_QUIRKS,
enable_quirks, gst_ks_video_src_class_init, gst_ks_video_src_init,
gst_ks_video_src_finalize, gst_ks_video_src_get_property,
gst_ks_video_src_set_property, gst_ks_video_src_reset,
gst_ks_video_src_apply_driver_quirks, gst_ks_video_src_change_state,
gst_ks_video_src_set_caps):
First driver quirk: work around Logitech's hostile driver software to
improve stability and performance. See comments for details.
Provide a property to disable driver quirks (enabled by default).
Be nice and specify G_PARAM_STATIC_STRINGS.
Remove unused dispose method.
Tweak include order.
Original commit message from CVS:
* configure.ac:
* sys/Makefile.am:
* sys/winks/Makefile.am:
* sys/winks/gstksclock.c:
* sys/winks/gstksclock.h:
* sys/winks/gstksvideodevice.c:
* sys/winks/gstksvideodevice.h:
* sys/winks/gstksvideosrc.c:
* sys/winks/gstksvideosrc.h:
* sys/winks/kshelpers.c:
* sys/winks/kshelpers.h:
* sys/winks/ksvideohelpers.c:
* sys/winks/ksvideohelpers.h:
New plugin for low-latency video capture on Windows (#519935).
Uses Kernel Streaming, the lowest level API for doing video capture
on Windows (more or less just raw ioctls).
