Original commit message from CVS:
2007-01-06 Andy Wingo <wingo@pobox.com>
* gst-libs/gst/audio/gstbaseaudiosink.c
(gst_base_audio_sink_class_init)
(gst_base_audio_sink_init):
(gst_base_audio_sink_activate_pull): Add an activate_pull function
to baseaudiosink, and tell basesink that we can work in pull mode.
This way the ring buffer thread drives the pipeline directly, if
pull mode is possible. There is some lingering nastiness regarding
capsnego, however.
(gst_base_audio_sink_callback): Implement the callback to pull
data. This interface is a bit light, though -- it should get a
GstFlowReturn return value at least.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_event), (gst_base_audio_sink_render):
* gst-libs/gst/audio/gstbaseaudiosink.h:
Make the clock sync code more accurate wrt resampling and playback
at different rates.
* gst-libs/gst/audio/gstringbuffer.c:
(gst_ring_buffer_commit_full), (gst_ring_buffer_commit):
* gst-libs/gst/audio/gstringbuffer.h:
Use better algorithm to interpolate sample rates.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_event), (gst_base_audio_sink_render):
* gst-libs/gst/audio/gstbaseaudiosink.h:
Extract rate from the NEWSEGMENT event.
Use commit_full to also take rate adjustment into account when writing
samples to the ringbuffer.
* gst-libs/gst/audio/gstringbuffer.c:
(gst_ring_buffer_commit_full), (gst_ring_buffer_commit),
(gst_ring_buffer_read):
* gst-libs/gst/audio/gstringbuffer.h:
Added _commit_full() to also take rate into account.
Use simple interpolation algorithm to resample audio.
API: gst_ring_buffer_commit_full()
* tests/examples/seek/scrubby.c: (speed_cb), (do_seek):
* tests/examples/seek/seek.c: (segment_done):
Don't try to seek with 0.0 rate, just pause instead.
Remove bogus debug line.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_setcaps):
Don't crash when ringbuffer is not yet created.
Patch by: Ville Syrjala <ville dot syrjala at movial dot fi>
Fixes#361634.
* gst/playback/gstplaybasebin.c: (new_decoded_pad_full):
* gst/playback/gststreamselector.c:
(gst_stream_selector_request_new_pad):
Activate pads befre adding them to running elements.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
When we have a timestamp, we can still perform clipping.
When we have no clock, we must play the sample ASAP.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
Add some more info in a WARNING.
* gst-libs/gst/audio/gstbaseaudiosrc.c:
(gst_base_audio_src_create):
Handle PAUSE in create function, use new -core addition to
wait for playing. Fixes pausing and resuming capture from an
audiosrc.
* gst-libs/gst/audio/gstringbuffer.c: (gst_ring_buffer_commit),
(gst_ring_buffer_read):
Constify some more.
Caller supports interrupted reads now.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_get_time), (gst_base_audio_sink_callback):
* gst-libs/gst/audio/gstbaseaudiosrc.c:
(gst_base_audio_src_get_time), (gst_base_audio_src_fixate),
(gst_base_audio_src_get_times), (gst_base_audio_src_get_offset),
(gst_base_audio_src_create), (gst_base_audio_src_change_state):
Do the delay calculation in the source/sink base classes as this is
specific for the capture/playback mode.
Try to fixate a bit better, like round depth up to a multiple of 8
bigger than width.
Handle underruns correctly by marking DISCONT on buffers and adjusting
timestamps to handle the gap.
Set offset/offset_end correctly on buffers.
* gst-libs/gst/audio/gstringbuffer.c: (gst_ring_buffer_pause),
(gst_ring_buffer_samples_done), (gst_ring_buffer_commit),
(gst_ring_buffer_read):
Remove resync and underrun recovery from the ringbuffer.
Fix ringbuffer read code on under/overrun.
Original commit message from CVS:
* configure.ac:
We require 0.10.10.1 now because of _wait_preroll().
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
Use gst_base_sink_wait_preroll().
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
Don't try to align a sample to an unknown value.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_provide_clock), (gst_base_audio_sink_render):
When the audio clock is slaved to another clock, never try to align
samples but trust the rate interpolation algorithm.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_class_init),
(gst_base_audio_sink_provide_clock):
Use gobject_class instead of G_OBJECT_CLASS (klass)
* gst-libs/gst/audio/gstbaseaudiosrc.c:
(gst_base_audio_src_class_init), (gst_base_audio_src_init),
(gst_base_audio_src_set_clock), (gst_base_audio_src_provide_clock),
(gst_base_audio_src_get_time),
(gst_base_audio_src_check_get_range), (gst_base_audio_src_create),
(gst_base_audio_src_create_ringbuffer):
Fix latency and buffer-time constants and properties ala basesink.
Implement pull based scheduling. Fixes#346527.
Set default blocksize in GstBaseSrc to 0, we default to pushing out
one segment.
Refuse slaving to another clock instead of silently not working.
Only provide a clock when we are actually able to do so.
Various small cleanups and compiler hints.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_class_init), (gst_base_audio_sink_setcaps):
* gst-libs/gst/audio/gstbaseaudiosink.h:
* gst-libs/gst/audio/gstringbuffer.c: (gst_ring_buffer_parse_caps),
(gst_ring_buffer_samples_done):
* gst-libs/gst/audio/gstringbuffer.h:
Document better the fact that latency_time and buffer_time are values
stored in microseconds, and not the usual GStreamer nanoseconds.
Change the variables (compatibly) that store them from GstClockTime
to guint64 to make it more clear that they're not storing clock times.
Also, remove the bogus property description that says the user can
specify -1 to get the default value, since that's never been the case.
When computing the default segment size for the ring buffer, make it
an integer number of samples.
When the sub-class indicates a delay greater than the number of
samples we've written return 0 from the audio sink get_time method.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_drain), (gst_base_audio_sink_event),
(gst_base_audio_sink_render), (gst_base_audio_sink_change_state):
patch to make timestamp checking more tollerant to rounding
errors given that real discontinuities are to be marked on
buffers. Fixes some asf files and #338778.
Also avoid some crashers when we receive an event in the
NULL state.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_event):
Starting the ringbuffer when we did not acquire it can cause
a deadlock, is pointless and causes nasty things for
subclasses.
Fixes gst-launch audiotestsrc num-buffers=0 ! alsasink.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_class_init), (gst_base_audio_sink_async_play),
(gst_base_audio_sink_do_play), (gst_base_audio_sink_change_state):
Implement new async_play vmethod to start slaving and allow
playback start in case of async PLAY state changes.
* gst-libs/gst/video/gstvideosink.c: (gst_video_sink_init):
Enable QoS with new method in base class.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_dispose):
Since we _parent the ringbuffer, we also need to
_unparent instead of a plain _unref.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_change_state):
* gst-libs/gst/audio/gstringbuffer.c: (wait_segment),
(gst_ring_buffer_may_start):
* gst-libs/gst/audio/gstringbuffer.h:
Only start playback if we are playing.
should fix#330748.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_provide_clock):
Don't try to provide a clock in the NULL state.
Original commit message from CVS:
* gst-libs/gst/audio/TODO:
Update TODO
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_get_offset):
When trying to play samples ASAP and we don't have a
previous sample, try to play at position 0 instead of
an invalid position.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_get_time), (gst_base_audio_sink_preroll),
(gst_base_audio_sink_render), (gst_base_audio_sink_change_state):
Always sync on first sample we receive when starting.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_get_time), (gst_base_audio_sink_setcaps),
(gst_base_audio_sink_drain), (gst_base_audio_sink_preroll),
(gst_base_audio_sink_render), (gst_base_audio_sink_change_state):
Use scale functions when possible.
Fix error messages.
Free clockid when after waiting for EOS.
Use G_(UN_)LIKLY when it makes sense.
Fix sample clipping bug found by Arwed v. Merkatz fixes#330789.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_provide_clock),
(gst_base_audio_sink_set_property),
(gst_base_audio_sink_get_property), (gst_base_audio_sink_render):
Don't try to provide a clock when we are not negotiated since
we might not be able to make it run.
Original commit message from CVS:
* gst-libs/gst/audio/TODO:
Updated.
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_drain), (gst_base_audio_sink_event):
On EOS, wait till the last sample is played before posting EOS.
Original commit message from CVS:
* gst-libs/gst/audio/gstaudiosink.c:
(gst_audioringbuffer_class_init), (gst_audioringbuffer_release),
(gst_audioringbuffer_pause):
Implement pause that does not wait for completion.
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render), (gst_base_audio_sink_change_state):
Don't drop buffers when going to PAUSED but perform preroll on
remaining samples now that core base class supports this.
* gst-libs/gst/audio/gstringbuffer.c: (gst_ring_buffer_release),
(gst_ring_buffer_pause_unlocked), (gst_ring_buffer_stop),
(gst_ring_buffer_commit):
Pause should not signal waiters.
Implement return value of _commit correctly.
Original commit message from CVS:
* gst-libs/gst/audio/audio.h: (GST_CLOCK_TIME_TO_FRAMES)
* gst-libs/gst/audio/gstbaseaudiosink.c: (gst_base_audio_sink_render)
use of gst_guint64_to_gdouble to be compliant with vs6
* gst/playback/gstdecodebin.c: (try_to_link_1)
* gst/videorate/videorate.c: (gst_video_rate_blank_data)
use of G_GINT64_CONSTANT for int64 constants
* win32/common/libgstinterfaces.def:
export some symbols (gst_mixer_get_type,gst_mixer_track_get_type)
* win32/vs6:
update and add new project files
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render), (gst_base_audio_sink_change_state):
Undo previous commit, it breaks resume after pause.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_setcaps), (gst_base_audio_sink_event),
(gst_base_audio_sink_preroll), (gst_base_audio_sink_render):
Improve debugging.
Post error when caps cannot be parsed.
Resync on discontinuity in the stream.
Clip samples to segment boundaries.
return WRONG_STATE sooner when we are flushing.
* gst-libs/gst/audio/gstbaseaudiosrc.c: (gst_base_audio_src_init),
(gst_base_audio_src_get_time), (gst_base_audio_src_create):
Make audiosrc operate in TIME.
Set TIMESTAMP and DURATION on buffers.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
Fix playback of non-synchronised streams by assuming a rate
of 1.0 instead of a random one.
Makes this work again:
gst-launch filesrc location=raw_audio.file ! 'audio/x-raw-int,
endianness=(int)4321, signed=(boolean)true, width=(int)16,
depth=(int)16, rate=(int)44100, channels=(int)2' ! audioconvert !
audioresample ! alsasink
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_class_init):
* gst-libs/gst/audio/gstbaseaudiosrc.c:
(gst_base_audio_src_class_init):
update strings, values are in microseconds
change the default sink buffer time to something that is smaller
(to help software volume mixing have a slightly lower delay) but
still be acceptable on Wim's laptop
Original commit message from CVS:
* gst-libs/gst/audio/TODO:
Updated TODO
* gst-libs/gst/audio/gstaudiosink.c:
(gst_audioringbuffer_open_device),
(gst_audioringbuffer_close_device), (gst_audioringbuffer_acquire),
(gst_audioringbuffer_release):
Small cleanups.
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_class_init), (gst_base_audio_sink_render),
(gst_base_audio_sink_change_state):
Slave to the master clock when going to PLAYING and unslave when
going to PAUSED.
* gst-libs/gst/audio/gstringbuffer.c:
(gst_ring_buffer_open_device), (gst_ring_buffer_close_device),
(gst_ring_buffer_acquire), (gst_ring_buffer_release),
(gst_ring_buffer_samples_done), (gst_ring_buffer_set_sample),
(gst_ring_buffer_clear_all), (wait_segment),
(gst_ring_buffer_commit), (gst_ring_buffer_read),
(gst_ring_buffer_advance):
* gst-libs/gst/audio/gstringbuffer.h:
Add some docs and cleanups.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_class_init), (gst_base_audio_sink_set_clock),
(gst_base_audio_sink_render), (gst_base_audio_sink_change_state):
Fix for calibration API change.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_class_init), (gst_base_audio_sink_set_clock),
(gst_base_audio_sink_render), (gst_base_audio_sink_change_state):
And we provide a clock by default, of course...
Original commit message from CVS:
* gst-libs/gst/audio/gstaudioclock.c: (gst_audio_clock_init):
This clock can be slaved to a master clock now.
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_class_init), (gst_base_audio_sink_init),
(gst_base_audio_sink_dispose), (gst_base_audio_sink_provide_clock),
(gst_base_audio_sink_set_clock),
(gst_base_audio_sink_set_property),
(gst_base_audio_sink_get_property), (gst_base_audio_sink_preroll),
(gst_base_audio_sink_render), (gst_base_audio_sink_change_state):
* gst-libs/gst/audio/gstbaseaudiosink.h:
Handle slaving the internal clock to the clock selected in the
pipeline.
Add property to make the basesink not provide a clock.
* gst-libs/gst/rtp/gstbasertpdepayload.c:
(gst_base_rtp_depayload_class_init), (gst_base_rtp_depayload_init),
(gst_base_rtp_depayload_wait):
* gst-libs/gst/rtp/gstbasertpdepayload.h:
We can use the clock in GstElement, no need to store it ourselves.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_provide_clock),
(gst_base_audio_sink_change_state):
Set ringbuffer to non-flushing when going to PAUSED, set to
flushing again when going to READY.
* gst-libs/gst/audio/gstringbuffer.c: (gst_ring_buffer_init),
(gst_ring_buffer_stop):
Start in flushing mode by default.
Don't set flushing in the _stop method, let the app call
this explicitly.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_event), (gst_base_audio_sink_get_offset),
(gst_base_audio_sink_render):
* gst-libs/gst/audio/gstringbuffer.c:
(gst_ring_buffer_open_device), (gst_ring_buffer_close_device),
(gst_ring_buffer_set_flushing), (gst_ring_buffer_start),
(gst_ring_buffer_pause_unlocked), (gst_ring_buffer_pause),
(gst_ring_buffer_stop), (wait_segment), (gst_ring_buffer_commit),
(gst_ring_buffer_read):
* gst-libs/gst/audio/gstringbuffer.h:
Add flushing mode to the ringbuffer so that it in all cases does
not try to handle more audio. This makes sure it does not try to
block anymore when flushing and fixes a livelock.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_get_offset), (gst_base_audio_sink_render):
Remove g_print
Use sync property from baseclass to disable sync.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_get_offset), (gst_base_audio_sink_render):
Buffers with no timestamps get aligned with previous buffers or
on underrun, played ASAP.
Original commit message from CVS:
* ext/theora/theoradec.c: (theora_dec_src_query),
(theora_dec_sink_event):
* ext/vorbis/vorbisdec.c: (vorbis_dec_src_query),
(vorbis_dec_sink_event), (vorbis_handle_identification_packet),
(vorbis_handle_data_packet):
* ext/vorbis/vorbisdec.h:
Fix old naming.
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
Don't try to sync on buffers without a timestamp.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
* gst-libs/gst/audio/gstbaseaudiosrc.c:
(gst_base_audio_src_create):
* gst-libs/gst/audio/gstringbuffer.c: (gst_ring_buffer_commit),
(gst_ring_buffer_read):
Cleanups.
Commit and read from ringbuffer in samples rather than bytes.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
Respect segment rate and accum when scheduling samples.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_event), (gst_base_audio_sink_render):
If we receive EOS we can start playback of what we had.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
* gst-libs/gst/audio/gstringbuffer.c: (gst_ring_buffer_commit):
* gst/volume/gstvolume.c: (gst_volume_class_init),
(volume_transform):
fixing lost sync, some more debugging
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
Resync if the buffer timestamps drift more than a 10th
of a second.
Original commit message from CVS:
2005-08-08 Andy Wingo <wingo@pobox.com>
* gst-libs/gst/audio/gstbaseaudiosink.c
(gst_base_audio_sink_change_state): Open the device in NULL->READY
like good elements should. Close on READY->NULL too.
* gst-libs/gst/audio/gstaudiosink.c
(gst_audioringbuffer_open_device,
(gst_audioringbuffer_close_device, gst_audioringbuffer_acquire)
(gst_audioringbuffer_release): Updates for new ring buffer API,
hook into the new audio sink api.
* gst-libs/gst/audio/gstaudiosink.h (GstAudioSinkClass.open)
(GstAudioSinkClass.close): Just open and close the device -- no
resource allocation or configuration.
(GstAudioSinkClass.prepare, GstAudioSinkClass.unprepare): New
vmethods, handle device setup and resource allocation.
* ext/alsa/gstalsasink.c (gst_alsasink_open, gst_alsasink_close)
(gst_alsasink_prepare, gst_alsasink_unprepare): Update for new
base class API.
* gst-libs/gst/audio/gstringbuffer.h
(GstRingBufferClass.open_device, GstRingBufferClass.close_device):
New vmethods.
* gst-libs/gst/audio/gstringbuffer.c (gst_ring_buffer_open_device)
(gst_ring_buffer_close_device, gst_ring_buffer_device_is_open):
New API functions. The device should be opened before acquiring
and closed after releasing.
Original commit message from CVS:
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_base_audio_sink_render):
Align samples even if we have roundoff errors in the
timestamp conversion.
Original commit message from CVS:
* gst-libs/gst/audio/gstaudiosink.c: (gst_audioringbuffer_init):
* gst-libs/gst/audio/gstbaseaudiosink.c:
(gst_baseaudiosink_class_init), (gst_baseaudiosink_dispose),
(gst_baseaudiosink_change_state):
* gst-libs/gst/audio/gstbaseaudiosink.h:
* gst-libs/gst/audio/gstringbuffer.c:
(gst_ringbuffer_set_callback):
Fix compilation error.
Ringbuffer starts out as not running.
Free our clock in dispose.
When releasing the ringbuffer we need to renegotiate so
clear the pad caps.
Original commit message from CVS:
Make ringbuffer faster and more simple by removing the locks
in the playback thread.
Add sample accurate playback based on buffer sample offsets.
Make the baseaudiosink provide a clock.
Parse caps in the base class.
Correctly handle seeking, flushing and state changes.
