This ensures that collectpads' cookie is properly updated so that when the streaming
threads will restart and be checking for the flushing status of all pads there will
be no inconsistent state.
This patch adds support for stationary white Gaussian noise.
The Box-Muller algorithm is used to generate pairs of independent
normally-distributed random numbers.
Fixes bug #586519.
When a seek failed upstream, make sure the adder sinkpad is set unflushing again
so that streaming can continue.
We only have a pending segment when we flushed.
Set the flush_stop_pending flag inside the appropriate locks and before we
attempt to perform the upstream seek.
Add some more comments.
Use the right lock to protect the flags in flush_stop.
See #585708
Set the target state of the newly added uridecodebins to somthing else that
PAUSED so that we keep their state in sync with the playsink state.
Fixes#585268
At least do the fix to sent the flush_stop when seeking failed to ensure we
keep no pads flushing. before it was send when the seeking worked which is just
plain wrong and was not the intention.
When no flush-stop has been sent by upstream, we have to send one ourselves to
continue playback. Do this as soon as the collect function is called instead of
after we possibly pushed segment events (that got then flushed out)
uridecodebin expects the passed connection-speed value in kbps, so we
need to divide the value stored in bps by 1000. Also, lower the upper
limit on the properties to the value that we can actually store in our
internal guint (which is plenty high enough)
Be even less restrictive in what we accept for .srt timestamps when
typefinding and parsing subrip subtitles and add a unit test for
the 'new' format. Fixes#585197.
When we are probing for streams, we want to set the queue size in such a way
that we can scan a maximum amount of data without consuming too much memory.
Therefore, remove the time limit on the queue and only stop scanning after 2MB
of data.
See #584104.
Recognise PGS subpicture streams and connect them to the SPU pad
in playsink. Unfortunately this fails badly with negotiation errors
if the SPU is not recent enough to support the stream. I'm not sure
how to add format negotiation in yet.
When using an audio sink without a "volume" property, volume control
would only work for the first song. For the next song, we'd try to
re-use the existing audio chain, but inadvertently set chain->volume
to NULL instead of to the existing volume element.
playbin2 inadvertently used autoaudiosink and autovideosink up to now,
since it would overwrite the sinks configured via the "audio-sink"
and "video-sink" properties with the stream-specific group sinks when
configuring the outputs. Those are usually NULL however, so that would
overwrite the configured sinks with NULL which makes playbin2 then
default to the auto sinks. Fix this by keeping a reference to each
configured sink in playbin2 and setting up the right sinks depending
on whether there is a stream-specific sink or not.
Fixes#584020.
Use two flags to remember volume/mute changes at times when we don't have the
audiochain yet (e.g. construction). Only set values when they were actualy
changed. This makes pulseaudio's stream restore functional.
Adder was relying that something else sends a flush stop. When using adder with
a livesource it was not getting a flush_stop and thus all pads downstream where
keept flushing. Mark a pending flush_stop and send it when we are working on
the new segment back in the streaming thread.
Add a queue2 after the raw output pads of certain sources such as those for uris
like cdda://
No tuning of the queue is done yet as the defaults seem to work fine for me.
Fixes#582528
The enum nick should be 'sine-table', not 'sine table'. Technically this is
an API/ABI change I guess, but anyone who was using this and didn't report
it deserves this.
Handle buffers with -1 timestamps better by keeping track of the en time of the
previous buffer and assuming the -1 timestamp buffer goes right after the
previous one.
when we have two buffers that are equally good, output the oldest buffer once to
minimize latency.
don't try to calculate latency when the input framerate is unknown.
Keep track of the autoplugged custom sinks and configure them in the playsink
element when we have collected all streams.
Also make sure that we only select one custom sink.
When unreffing the internal sink, we don't need to change the state to NULL.
mp3_type_find could suggest already when only a single valid header
was found, if it ran out of data before the end of the next frame.
Therefore, ignore the last found frame if it was incomplete.
Fixes bug #579692.
Make playsink go async to the PAUSED state instead of relying on uridecodebin
for async behaviour in playbin. This solves some problems (mainly with DVD)
where the pipeline would go to PLAYING before preroll completed, failing to
select the audiosink clock.
Fixes#581727
When calculating the input/output buffer sizes in the transform_size function,
take the number of channels into account, so we don't end up calculating
a buffer size that only contains a partial number of audio frames.
Also, when going from output size to input size, round down rather than
up, so as to calculate the minimum number of samples that *might* yield
a buffer of the intended destination size.
Fixes: #580470 and #580952
When videorate duplicates a buffer with a DISCONT flag, it copies the discont on
the first pushed buffer but fails to clear it for subsequent buffers. This
causes theoraenc!oggmux and possibly other elements to consider this a discont
stream.
Fix videorate to produce discont as the first buffer and after a flushing seek.
Fixes#580271.
The 2s limit is way too small for a lot of files (which have an interleave
in time of between 3 and 5s). Instead, leave it to the initial 5s value
and reduce the other limits (allowing us to stay memory-efficient).
First check the pad caps if they are raw before setting the raw_decoding_mode to
TRUE. Fixes playback of transport streams and other streams that require large
queues.
Fixes#579734
Adds a new property in multifdsink, resend-streamheader.
If this property is false, the multifdsink will not send the streamheader if
there's already one set for a particular client.
There are some formats in which every stream needs to start with a certain
blob, but you can't inject this blob at leisure. If the producer wants to
change the blob in question and sets in as the streamheader on the outgoing
buffers' caps, new clients of multifdsink will get the new streamheader, but
old clients will break, because they'll see the blob in the middle of the
stream.
The property is true by default, so existing code will not see any difference.
Fixes#578118.
Add a property to disable listening to client writes. This property is usefull
when other code will deal with reading from the client socket.
API: GstMultiFdSink::handle-read property
Clear the target of our ghostpads before we remove the pad from the element.
This to make sure that the internal pad is not left linked to whatever pad we
were ghosted to. This should only be a problem when we leak the ghostpads.
Also release our subpicture pads.
Fixes#577288.
Raw decoding mode removes almost all buffering in video and audio queues
when a source providing already decoded video/audio is detected, on the
possibly bogus assumption that such a source should provide sufficient
internal queueing. Fixes playback on some DVDs, and improves it
on all.
If one side has a preference for a particular sample rate or set of sample rates, we
should honour this in the caps we advertise and transform to and from, so that elements
actually know about the other side's sample rate preference and can negotiate to it
if supported. Also add unit test for this.
This prevents valgrind warnings when accessing the "x" parts
of xRGB and friends in other elements that handle (and can handle)
xRGB like ARGB (for example videoscale).
When reusing playbin with visualisations, reset the async property on the video
sink because some sinks might dynamically recreate their sinks.
Fixes#576188
When we have the textpad configured, enable and disable the subtitles by setting
the silent flag on the overlay element instead of trying to remove elements.
See #576187
Updated the examples in the README to actually work. Add them to api docs. Tests
the api-docs and fix the section names to make the docs actualy show up.
The example for "tcpserversrc" needs review (might be an element bug).
Link after doing the state change and unlink before shutting down. Makes the
window for causing races in toggling the visualisations smaller.
See #576187.
Remove the group GCond that we used for waiting for groups to finish because we
use pad blocking on the selectors and counters instead for waiting for the
groups to complete.
remove the obsolete about_to_finish variable set while emiting the
about-to-finish signal and fix some old comments.
We don't need to take the playbin lock when querying the uridecodebin.
When we make a group connected to a demuxer, keep an extra dynamic refcount for
the group which is only decremented when no_more_pads or a multiqueue overrun is
detected. This way we avoid a race between exposing the group while more dynamic
refs are added from new pads.
Fixes#575588.
Sync the state of the newly added chains to the state of the parent sink element
to avoid lost async-start messages. Fixes cdda:// async-done message storm.
When streams are not selected in the selector, return NOT_LINKED so that
upstream elements can skip decoding. Only do this for audio and video pads
because for text streams the overhead is smaller and they could come from
external files.
Set the custom sink async=FALSE to not make it participate in preroll because we
are dealing with sparse streams.
Try to set sync=TRUE on the custom text sink.
Release the shutdown lock when we wait for other groups to complete or else we
have a deadlock when the other group completes and tries to grab the shutdown
lock.
Fixes#575550.
The flac frame header typefinder overstates the likelihood of a match, leading
to false positives with e.g. aac streams and PDF files. Reduce probabilty
returned from LIKELY to POSSIBLE for the frame header matchin code.
Fixes#574939.
Detect more variations and also bail out in more cases where the values
don't make sense. Furthermore, add width/height and bpp to the caps,
because we can.
Add property to playbin2 to configure a custom sink that receives the raw
subtitle buffers instead of using a textoverlay.
Improve the property finding code to make it more usable.
Use property find code to find async properties in custom sinks that are bins.
Improve text overlay code to gracefully handle missing elements.
Use scan context for initial peek as well. Peek 6 bytes in the initial
peek rather than 5 bytes, to match the length of the memcmp we're doing
on that data later. Return immediately when we found caps from looking
at the beginning of the data - no point in continuing to scan the next
64kB for something matching a frame header.
Disconnect the notify::caps signal in our callback (it'll be re-added
if we're not, in fact, finished getting complete caps). Ensures that
caps changes mid-stream (e.g. from an mp3 that changes from
stereo->mono mid-file) don't cause us to try to add a new pad.
Make it possible to request a flushing pad from the playsink. We can eventually
use these flushing pads to quickly terminate the dataflow when we are shutting
down.
Release the group lock while we perform the state changes on the uridecodebins
because that might trigger callbacks that we need to handle with the group lock
taken. Avoids a possible deadly embrace in some id3/flac files.
Fixes#567396.
When setting the quality/filter-length while PLAYING the
resampling context will be destroyed and created again in
some cases, which will cause crashes in the transform function
if it's called at that time.
Rather than only checking for volume property on the audio sink
directly, recursively look for it on sinks within it (if it's a bin).
Allows use of sink-as-volume-control where the application has supplied
an audio-sink bin that includes a real audio sink internally.
Don't keep extra references to volume and mute elements; we don't need
to do so.
Ensure we unref pads that we have references to, and release request
pads.
Because core now supports typefindfactories without a typefind function we can
register a factory fo GSM that will --if all else fails-- assume the file is a
GSM file based on the registered extension.
Fixes#566661.
We can use gst_element_link_pads() instead of the more generic
gst_element_link() function because we know the pads. This saves some cycles
because the more generic function needs to search for possible compatible caps
etc.
Original commit message from CVS:
* gst/playback/gstplaybin2.c: (gst_play_bin_class_init),
(gst_play_bin_set_uri), (gst_play_bin_set_suburi),
(no_more_pads_cb), (drained_cb), (group_set_locked_state_unlocked),
(activate_group), (deactivate_group), (groups_set_locked_state),
(gst_play_bin_change_state):
Fix some comments and docs.
Post an error message when we fail to link the selector to the sink.
Remove pushing of EOS, this seems unneeded.
Lock the state of deactivated groups so that they don't accidentally
reactivate when the playbin2 state changes.
Reuse uridecodebins.
Unlock and relock state of groups when playbin goes to NULL.
Fixes#566654.
Fixes#566341.
* gst/playback/gsturidecodebin.c: (pad_removed_cb), (type_found):
Only do something in the pad removed callback when we are dealing with
our sourcepads because the sinkpads don't have a ghostpad.
Original commit message from CVS:
* gst/playback/gstplaybin2.c: (notify_source_cb), (activate_group):
Disconnect signal handlers before destroying a previous decodebin so
that we don't end up causing deadlocks. Fixes#566586.
Original commit message from CVS:
* gst/audiotestsrc/gstaudiotestsrc.c:
(gst_audio_test_src_class_init), (gst_audio_test_src_init),
(gst_audio_test_src_check_get_range),
(gst_audio_test_src_set_property),
(gst_audio_test_src_get_property):
* gst/audiotestsrc/gstaudiotestsrc.h:
Add property to control pull/push based scheduling.
Original commit message from CVS:
* gst/playback/gstplaybin2.c: (no_more_pads_cb):
Add some debug info.
* gst/playback/gstplaysink.c: (gen_video_chain), (gen_audio_chain),
(gst_play_sink_reconfigure), (gst_play_sink_request_pad),
(gst_play_sink_release_pad):
Add some more debug info.
Reconfigure the audio chain when we switch between raw and encoded audio
in gapless playback.
Original commit message from CVS:
* gst/adder/Makefile.am:
* gst/adder/gstadder.c:
Cleanup variable names to make the adder-loop easier to understand.
Also try to use liboil to spee it up, but ifdef it out as it does not
make any change for me (Intel pentim M (sse,sse2) please try on other
systems).
Original commit message from CVS:
* docs/plugins/Makefile.am:
* docs/plugins/gst-plugins-base-plugins-docs.sgml:
* docs/plugins/gst-plugins-base-plugins-sections.txt:
* gst/tcp/gsttcpclientsink.c:
* gst/tcp/gsttcpclientsrc.c:
* gst/tcp/gsttcpserversrc.c:
Add minimal docs to make the remaining tcp elements show up.
Fixes#564139.
Original commit message from CVS:
* gst/playback/gsturidecodebin.c: (gst_uri_decode_bin_finalize):
Free the factory array when finalizing.
* sys/v4l/gstv4lsrc.c: (gst_v4lsrc_base_init):
Use a GstStaticPadTemplate since the src pad caps are fixed.
Original commit message from CVS:
* gst/subparse/samiparse.c: (sami_context_push_state),
(sami_context_pop_state), (start_sami_element), (end_sami_element):
Some versions of libxml seem to be very picky as to strict formatting
of the input and never 'close' the final </body> tag.
In order to fix that bad behaviour, we trigger the flushing of
remaining data on both </body> and </sami>.
Fixes#557365
Original commit message from CVS:
Patch by: Guillaume Emont <guillaume at fluendo dot com>
* gst/typefind/gsttypefindfunctions.c: (plugin_init):
Add typefinders for MS Word files and OS X .DS_Store files to
prevent them to be recognized as MPEG files. Fixes bug #564098.
Original commit message from CVS:
* gst/playback/gstplaysink.c: (gen_audio_chain),
(gst_play_sink_reconfigure):
Add some more debug info.
Fix linking of just an encoded sink.
Handle failure to create a sink chain more gracefully than crashing.
Original commit message from CVS:
* gst/playback/gstplaybin2.c: (pad_added_cb):
Error out with a missing-plugin error when the input-selector was not
found.
* gst/playback/gstplaysink.c: (gst_play_sink_reconfigure):
Indentation.
Original commit message from CVS:
* gst/playback/gstplaysink.c: (gst_play_sink_class_init),
(gst_play_sink_dispose), (gst_play_sink_finalize), (try_element),
(gen_video_chain), (gen_audio_chain), (gst_play_sink_reconfigure),
(gst_play_sink_send_event), (gst_play_sink_change_state):
Use G_DEFINE_TYPE.
Try to set the selected sink to READY before using it. This will allow
for detection of incompatible formats sooner.
Don't cause a fatal error when conversion elements are missing but post
a missing-element message and a warning instead because things might
still link and run fine.
Simplyfy the construction of audio and video sink chains.
Original commit message from CVS:
Patch by: Luis Menina <liberforce at freeside dot fr>
* gst-libs/gst/floatcast/floatcast.h:
* gst/typefind/gsttypefindfunctions.c:
Include glib.h instead of a specific GLib header. Including single
GLib headers is deprecated. Fixes bug #563904.
Original commit message from CVS:
* gst/volume/gstvolume.c: (gst_volume_class_init),
(volume_before_transform), (volume_transform_ip):
Use new basetransform vmethod to reconfigure the dynamic properties and
any pending volume/mute changes. Fixes#563508.
Original commit message from CVS:
* gst/playback/gstdecodebin.c:
* gst/playback/gstdecodebin2.c:
Add basic docs to decodebin and link to decodebin from decodebin2.
Original commit message from CVS:
* docs/plugins/gst-plugins-base-plugins.args:
* docs/plugins/gst-plugins-base-plugins.hierarchy:
* docs/plugins/gst-plugins-base-plugins.interfaces:
* docs/plugins/gst-plugins-base-plugins.prerequisites:
* docs/plugins/inspect/plugin-videorate.xml:
* gst/speexresample/gstspeexresample.c:
Update documentation of speexresample for the new element name.
Original commit message from CVS:
* gst/speexresample/gstspeexresample.c: (plugin_init):
Update the debug category from speex_resample to audioresample.
Original commit message from CVS:
* gst/playback/gstplaybin2.c:
Add notification of current stream. Add ability to configure buffer
sizes.
* gst/playback/gsturidecodebin.c:
Add ability to configure buffer sizes for streaming mode.
Bug #561734.
Original commit message from CVS:
* gst/playback/gstplaysink.c: (gen_audio_chain):
Don't post an error when we can't configure the volume but post a
warning instead. Fixes#561780.
Original commit message from CVS:
Patch by: Jonathan Rosser <jonathan.rosser@rd.bbc.co.uk>
* gst/videotestsrc/gstvideotestsrc.c:
* gst/videotestsrc/gstvideotestsrc.h:
* gst/videotestsrc/videotestsrc.c:
* gst/videotestsrc/videotestsrc.h:
Add a zone plate pattern generator based on BBC R&D Report
1978/23 (yeah *that* 1978). Try 'videotestsrc pattern=zone-plate
kx2=20 ky2=20 kt=1'.
Original commit message from CVS:
* gst/speexresample/gstspeexresample.c:
(gst_speex_resample_class_init), (gst_speex_resample_set_property),
(gst_speex_resample_get_property):
Add a "filter-length" property that maps to the quality values
for compatibilty with audioresample.
Original commit message from CVS:
* gst/playback/gstdecodebin2.c:
If the top-level type of the stream is plain text, don't try to decode
it, matching behaviour of decodebin.
* gst/playback/gstplaysink.c:
If we fail to generate a text chain (e.g. due to missing optional
plugins), don't crash.
Original commit message from CVS:
* gst/videotestsrc/gstvideotestsrc.c:
* gst/videotestsrc/gstvideotestsrc.h:
* gst/videotestsrc/videotestsrc.c:
* gst/videotestsrc/videotestsrc.h:
Add "colorspec" property, specifying whether to generate BT.601
or BT.709 video. This only affects YCbCr values, not RGB, since
if you're generating a 709 test pattern, presumably you want
709 RGB primaries, not 601. Also add "smpte75" pattern, which
uses 75% colors instead of 100%, since this is often more useful
for testing (and also follows the SMPTE EG-1 guideline).
Original commit message from CVS:
* gst/audioresample/gstaudioresample.c:
Guard against a NULL dereference I somehow encountered -
with a FLUSH_STOP arriving either before basetransform _start(),
or after _stop().
* gst/typefind/gsttypefindfunctions.c:
Make sure we never jump backwards when typefinding corrupt mov files.
Original commit message from CVS:
* gst/typefind/gsttypefindfunctions.c: (jp2_type_find),
(plugin_init):
Improve typefinding of ISO JPEG2000 mime types.
Original commit message from CVS:
* gst/playback/gstplaybin2.c: (deactivate_group):
don't try to unlink the selector sinkpad when we don't have it yet. This
can happen if an error occured before the group was complete.
Original commit message from CVS:
* gst/playback/gstplaybin2.c: (activate_group):
Catch state change errors and stop from the uridecodebin elements
instead of trying to continue in vain.
Original commit message from CVS:
* gst/playback/gstplaysink.c: (gst_play_sink_reconfigure):
Don't try to do crazy things when we only have a text pad without a
video pad. Fixes#559478.
Original commit message from CVS:
* gst/volume/gstvolume.c: (volume_update_real_volume),
(gst_volume_set_volume), (gst_volume_get_volume),
(gst_volume_set_mute), (gst_volume_init), (volume_setup),
(volume_transform_ip), (volume_update_mute),
(volume_update_volume), (volume_get_property):
* gst/volume/gstvolume.h:
Keep negotiated state in a separate variable.
Protect the volume and mute properties with the object lock.
Protect modifying the transform with the transform lock.
Original commit message from CVS:
* gst/ffmpegcolorspace/gstffmpegcodecmap.c:
(gst_ffmpeg_pixfmt_to_caps):
Only convert caps to string when debug is enabled.
Original commit message from CVS:
* gst/speexresample/gstspeexresample.c:
(gst_speex_resample_convert_buffer):
Add TODO at the top of the file for enabling SSE/ARM specific
optimizations and choosing the fastest implementation at runtime.
Add g_assert_not_reached() at two places that should really never
be reached.
Original commit message from CVS:
* gst/speexresample/gstspeexresample.c:
(gst_speex_resample_check_discont):
Fix format string and arguments.
* gst/speexresample/resample_sse.h:
Add missing file.
Original commit message from CVS:
* gst/speexresample/Makefile.am:
* gst/speexresample/gstspeexresample.c:
(gst_speex_resample_base_init), (gst_speex_resample_get_funcs),
(gst_speex_resample_convert_buffer), (_benchmark_int_float),
(_benchmark_int_int), (_benchmark_integer_resampling),
(plugin_init):
* gst/speexresample/gstspeexresample.h:
* gst/speexresample/resample.c:
* gst/speexresample/speex_resampler_double.c:
* gst/speexresample/speex_resampler_float.c:
* gst/speexresample/speex_resampler_int.c:
* gst/speexresample/speex_resampler_wrapper.h:
Add missing headers to Makefile.am.
Update copyright, years and my mail address.
Benchmark the integer resampling implementation against the
float implementation and use the faster one for 8/16 bit integer
input. On most recent systems the floating point version is faster.
Original commit message from CVS:
* gst/speexresample/gstspeexresample.c:
(gst_speex_resample_convert_buffer):
The length for the buffer conversion function is the number of
audio frames, i.e. we need to multiply it by the number of channels
to get the number of values. Also spotted by the unit test after
running in valgrind.
Original commit message from CVS:
* tests/check/elements/speexresample.c: (element_message_cb),
(eos_message_cb), (test_pipeline), (GST_START_TEST),
(speexresample_suite):
Add pipeline unit tests for testing all supported formats with
up/downsampling and different in/outrates.
* gst/speexresample/gstspeexresample.c:
(gst_speex_resample_push_drain), (gst_speex_resample_process):
* gst/speexresample/speex_resampler_wrapper.h:
Fix bugs identified by the testsuite.
