For reverse playback it is important to handle correctly the frame sync
points, which is set when the input buffer doesn't have the DELTA_UNIT flag.
This is handled correctly when decoder is packetized, but when it is not the
frame's sync point is not copied, and the reverse playback never decodes frame
batches.
The current patch adds the buffer's flags to the Timestamp list, where the
timestamp and duration of the input buffers are hold.
There were two consecutive log messages in gst_video_decoder_decode_frame().
Given the information they provide, it is more efficient to squash them into a
single one.
The playback rate is hold in the input_segment member variable, not in the
output_segment, and the parse_gather list was never filled because of that.
This patch changes the comparison with input_segment.
The output segment is only set up after data is output, which might be far in
the future for reverse playback. Also we are here interested in the state at
the current *input* frame (which is the keyframe), not any possible output.
For reverse playback the same behaviour was already implemented in
flush_parse().
For reverse playback, chain_forward() is only used to gather frames and not
for decoding, and it is actually called by the draining logic, causing an
infinite recursion.
While it's a bit tricky to discard frames *before* decoding (because
we might not be sure which data is needed or not by the decoder), we
can discard them after decoding if they are too late anyway.
Any following basetransform based element or similar would drop the frame too.
When asked to just decode keyframe, if we got a keyframe drain out
the decoder straight away.
This avoids having to wait for the next frame and reduces delay even
more.
https://bugzilla.gnome.org/show_bug.cgi?id=767232
This ensures the decoder is properly drained out when receiving a
DISCONT buffer. The optimal way of doing this would have been to
receive a GAP event before hand but it is not always possible.
Fixes big delays with some decoders (ex gst-libav) that will not
drain out data when only decoding keyframes.
https://bugzilla.gnome.org/show_bug.cgi?id=767232
gst_buffer_copy_region() does not copy the timestamp if it doesn't start
with the first byte. We just skip the tag here, so the timestamp is still
valid.
https://bugzilla.gnome.org/show_bug.cgi?id=767173
The base class was setting the DISCONT flag before checking whether the buffer
would be in segment or not.
Fix issues with DISCONT flags not being properly propagated downstream when
decoders buffers were out of segment.
https://bugzilla.gnome.org/show_bug.cgi?id=766800
gst_video_sink_center_rect() can be called without a GstVideoSink
having been instantiated so we can't relly on the video sink
class_init function to init the category.
Fix a warning when running:
GST_CHECKS=test_video_center_rect GST_DEBUG=6 G_DEBUG=fatal_warnings make libs/video.check-norepeat
https://bugzilla.gnome.org/show_bug.cgi?id=766510
If we e.g. have AVI with DV container with video/audio inside the DV
container, we can't handle this at this point with an encoding profile.
Instead of erroring out, flatten the container hierarchy.
https://bugzilla.gnome.org/show_bug.cgi?id=765708
This tag match the EXIF_TAG_FOCAL_LENGTH_IN_35_MM_FILM exif tag and is
stored on a short. Hence there is a precision loss compared to the
GstTag which is a double value.
https://bugzilla.gnome.org/show_bug.cgi?id=753930
It is the 35 mm equivalent focal length of the lens, mainly used in
photography. Tag value is stored in a double value to be consistent with
GST_TAG_CAPTURING_FOCAL_LENGTH.
https://bugzilla.gnome.org/show_bug.cgi?id=753930
When converting discoverer output to an encoding profile, it makes sense to
omit these. It's very very unlikely that our encoder is going to produce bit
by bit the same codec_data or streamheader.
https://bugzilla.gnome.org/show_bug.cgi?id=765534
We add a couple of new functions gst_sdp_media_parse_keymgmt and
gst_sdp_media_parse_keymgmt. We also implement
gst_sdp_message_attributes_to_caps and gst_sdp_media_attributes_to_caps
in terms of these new functions and also gst_mikey_message_to_caps.
Current implementation requires all srtp and srtcp parameters to be
given in the caps. MIKEY uses only one algorithm for encryption and one
for authentication so we now allow passing srtp or srtcp parameters. If
both are given srtp parametres will be preferred.
https://bugzilla.gnome.org/show_bug.cgi?id=765027
As said in its doc GST_AUDIO_CHANNEL_POSITION_NONE is meant to be used
for "position-less channels, e.g. from a sound card that records 1024
channels; mutually exclusive with any other channel position".
But at the moment using such positions would raise a
'g_return_if_reached' warning as gst_audio_get_channel_reorder_map()
would reject it.
Fix this by preventing any attempt to reorder in such case as that's not
what we want anyway.
https://bugzilla.gnome.org/show_bug.cgi?id=763799
We currently don't log much about channel positions making debugging
harder as it should be. This is the first step in my attempt to improve
this.
https://bugzilla.gnome.org/show_bug.cgi?id=763985
There is a small window of time where the audio ringbuffer thread
can access the parent thread variable, before it's initialized
by the parent thread. The patch replaces this variable use by
g_thread_self().
https://bugzilla.gnome.org/show_bug.cgi?id=764865
Since the allocation query caps contains memory size and the pad's caps
contains the display size, an audio encoder or decoder might need to allocate
a different buffer size than the size negotiated in the caps.
This patch splits this logic distinction for audiodecoder and audioencoder.
Thus the user, if needs a different allocation caps, should set it through
gst_audio_{encoder,decoder}_set_allocation_cap() before calling the negotiate()
vmethod. Otherwise the allocation_caps will be the same as the caps in the
src pad.
https://bugzilla.gnome.org/show_bug.cgi?id=764421
Since the allocation query caps contains memory size and the pad's caps
contains the display size, a video encoder or decoder might need to allocate
a different frame size than the size negotiated in the caps.
This patch splits this logic distinction for videodecoder and videoencoder.
The user if needs a different allocation caps, should set the allocation_caps
in the GstVideoCodecState before calling negotiate() vmethod. Otherwise the
allocation_caps will be the same as the caps set in the src pad.
https://bugzilla.gnome.org/show_bug.cgi?id=764421
Doing so prevents us dropping buffers in the rare, but possible, situations,
when the stream changes SSRC and new sequence numbers does not differ
much from the last sequence number from previous SSRC. For example:
ssrc - 0xaaaa 101,102,103,104 ssrc - 0xbbbb 102, 103, 104, 105...
In the scenario above we don't want to drop the first 3 packets of
0xbbbb stream.
https://bugzilla.gnome.org/show_bug.cgi?id=764459
The function gst_discoverer_info_copy doesn't copy the data members seekable
and result of the source GstDiscovererInfo.
In the case of copying a GstDiscovererInfo for later use, the seekbale will be
undefined, which in practice usually will be false, even though the seekable of
the original GstDiscovererInfo is true.
https://bugzilla.gnome.org/show_bug.cgi?id=762710
The parameter type was wrongly documenting that a GstVideoInfo structure
pointer was needed, while it needs a GstVideoFormatInfo structure
pointer.
https://bugzilla.gnome.org/show_bug.cgi?id=764414
P010 is a YUV420 format with an interleaved U-V plane and 2-bytes per
component with the the color value stored in the 10 most significant
bits.
https://bugzilla.gnome.org/show_bug.cgi?id=761607
---
Changes since v2:
- Set bits=16 in DPTH10_10_10_HI
Changes since v1:
- Fixed x-offset calculation in uv.
- Added 6-bit shifts to FormatInfo.
Store the filter in the desired sample format so that we can simply do a
linear or cubic interpolation to get the new filter instead of having to
go through gdouble and then convert.
Remove some unused variables from the inner product functions.
Make filter coefficients by interpolating if required.
Rename some fields.
Try hard to not recalculate filters when just chaging the rate.
Add more proprties to audioresample.
Rearrange the oversampled taps in memory to make it easier to use
SIMD instructions on them. this simplifies some sse code.
Add some more optimizations
Improve int16 resampling by using pmaddwd
Use intrinsics to scale and pack int16 samples
Align the coefficients so that we can use aligned loads
Add padding to taps and samples so that we don't have to use partial
loads for the remainder of the loops.
Remove copy_n, we can reuse the plain copy function with some new
parameters.
Align and pad the sample array.
Remove the consumed/produced output fields from the resampler and
converter. Let the caler specify the right number of input/output
samples so we can be more optimal.
Use just one function to update the converter configuration.
Simplify some things internally.
Make it possible to use writable input as temp space in audioconvert.
Base class type name should not reference libvisual since not all child
elements use this. This was an oversight when merging audiovisualizers into
a common base class.
libgstreamer currently exports some debug category
symbols GST_CAT_*, but those are not declared in any
public headers.
Some plugins and libgstvideo just use GST_DEBUG_CATEGORY_EXTERN()
to declare and use those, but that's just not right at
all, and it won't work on Windows with MSVC. Instead look
up the categories via the API.
If we don't have writable memory, make sure to make a copy of the input
samples into a temporary (writable) buffer, even if we are dealing with
a native intermediate format that we don't need to call the unpack
function for.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=761655
gst_pad_get_allowed_caps() will return NULL if the srcpad has no peer.
In that case, use gst_pad_peer_query_caps() with template caps as filter
to have negotiated output caps properly before forwarding GAP event.
https://bugzilla.gnome.org/show_bug.cgi?id=761218
Add missing ':' to tile_ws and tile_hs fields documentation to avoid
bad render of these two fields, mark reserved bytes as private to hide
field and avoid gtkdoc warning and add parameters description to
documented macro to avoid gtkdoc warnings.
https://bugzilla.gnome.org/show_bug.cgi?id=761132
In gst_video_info_to_caps(), make sure we end up with an RGB matrix for
RGB formats and warn when the GstVideoInfo colorimetry is wrong.
In gst_video_info_from_caps(), fix the GstVideoInfo with an RGB matrix
for RGB formats and warn about inconsistent caps.
See https://bugzilla.gnome.org/show_bug.cgi?id=759624
For RGB formats, the matrix in the colorimetry (conversion from YUV to
RGB) is irrelevant and we should ignore it and assume the identity
transform for everything we do.
Fixes https://bugzilla.gnome.org/show_bug.cgi?id=759624
It's quite unexpected behaviour that various subclass settings are just
reset before set_format(). Unfortunately changing this now has the risk
of breaking existing code but we should reconsider this for 2.0.
When the input and output formats are the same and in a possible
intermediate format, avoid unpack and pack.
Never do passthrough channel mixing.
Only do dithering and noise shaping in S32 format
Add support for float and int16 mixing
Remove in-place processing, this simplifies things as we won't be using it.
Don't do clipping for float audio formats
Process as many samples as we can from the input and return the number
of processed samples from the chain. This simplifies some code.
Fix the IN_WRITABLE handling, don't overwrite the flags.
Pass flags in _converter_new() so that we can configure ourselves
differently depending on some options.
SOURCE_WRITABLE -> IN_WRITABLE because the array is called 'in'
Simplify the API, we don't need the consumed and produced output
arguments. The caller needs to use the _get_in_frames/get_out_frames API
to check how much input is needed and how much output will be produced.
When we run the loop for another time and do not have a preset name, we would
try to print the preset name of a previous iteration that is already freed.
Also move some other variables into the block where they are actually used
to prevent similar mistakes in the future.
CID 1346536
We did not take the sample size into account. Rearrange the tests to have more
conversion test and an extra test case for passthrough operations.
Fixes#759890
Rename samples to num_samples, since we also have samples in chain, but that is
the data pointer. Always use gzize for num_samples. Make the log output a bit
more homogenous.
The EOS event can be propagated to the downstream elements when
is_eos flag remains set even after leaving the flushing state.
This fix allows this element to normally restart the streaming
after receiving the flush event by clearing the is_eos flag.
https://bugzilla.gnome.org/show_bug.cgi?id=759110
Rework the main processing loop. We now create an audio processing
chain from small core functions. This is very similar to how the
video-converter core works and allows us to statically calculate an
optimal allocation strategy for all possible combinations of operations.
Make sure we support non-interleaved data everywhere.
Add functions to calculate in and out frames and latency.
Match the section 5.1.2 of the HTTP/1.0 spec by using relative URIs unless we
are using a proxy server. Also, send Host header for compatability with
HTTP/1.1 and some HTTP/1.0 servers.
https://bugzilla.gnome.org/show_bug.cgi?id=758922
gst_rtsp_connection_connect_with_response accepts a response pointer
which it fills with the response from setup_tunneling if the
connection is configured to be tunneled. The motivation for this is to
allow the caller to inspect the response header to determine if
additional authentication is required so that the connection can be
retried with the appropriate authentication headers.
The function prototype of gst_rtsp_connection_connect has been
preserved for compatability with existing code and wraps
gst_rtsp_connection_connect_with_response.
https://bugzilla.gnome.org/show_bug.cgi?id=749596
Depayloaders will look at rtpbuffer->buffer for the discont flag.
When we set the discont flag on a buffer in the rtp base depayloader
and we have to make the buffer writable, make sure the rtpbuffer
actually contains the newly-flagged buffer, not the original input
buffer. This was introduced with the addition of the process_rtp_packet
vfunc, but would only trigger if the input buffer wasn't flagged
already and was not writable already.
When we detect a discont and the input buffer isn't already flagged
as discont, handle_buffer() does a gst_buffer_make_writable() on the
input buffer in order to set the flag. This assumed it had ownership
of the input buffer though, which it didn't. This would still work
fine in most scenarios, but could lead to crashes or mini object
unref criticals in some cases when a discont is detected, e.g. when
using pcapparse in front of a depayloader. This problem was
introduced in bc14cdf529.
In the case where the stream doesn't have a framerate set and the frames
don't have a duration set, we still want to use the clipping path to
make sure we don't push buffers outside of the segment.
The problem was the previous iteration was setting a duration of 2s, which
meant that any buffer which was less than 2s before the segment start would
end up getting pushed.
Instead, use a saner 40ms (25fps single frame duration) to figure out whether
the frame could be within the segment or not
Some servers incorrectly parse header names with strict case-sensitivity. For
compatibility with these systems change X-Sessioncookie to x-sessioncookie.
https://bugzilla.gnome.org/show_bug.cgi?id=758921
Any latency query before this will not get the correct latency so a new
latency query should be triggered once the audio sink know its own latency.
Without this the initial latency query from the pipeline arrives too early
sometimes and the resulting latency is too short.
https://bugzilla.gnome.org/show_bug.cgi?id=758911
Some Opus files found on the wild have 0 in the version field of the
OpusHead header, instead of the correct value of 1. The files still
play, don't make this error fatal.
https://bugzilla.gnome.org/show_bug.cgi?id=758754
Commit ff6d1a2a25 changed sample's type from
gint to gsize (and renamed it to in_samples). gsize is an unsigned long,
which means it can never be a negative value and the check making sure that
in_samples is >= 0 is never going to be false. Removing it.
CID 1338689
Move the audio quantize code from audioconvert to the audio library.
work on making an audio converter helper function similar to the video
converter.
Fold fastrandom directly into the quantizer, add some ORC code to
optimize this later.
Rename _get_default_mask() to _get_fallback_mask() to make it more
clear that the function only provides a fallback if nothing else can be
done. Also clarify this in the documentation.
API: gst_audio_channel_get_fallback_mask()
Add a TRUNCATE_RANGE flag for unpack functions to fill the least
significate bits with 0 (as did the old code). Also add functions
that don't truncate. Use the TRUNC flag in audioconvert for
backwards compatibility for now.
No need to use G_GINT64_FORMAT for potentially negative values of
GstClockTimeDiff. Since 1.6 these can be handled with GST_STIME_ARGS.
Plus it creates more readable values in the logs.
https://bugzilla.gnome.org/show_bug.cgi?id=757480
