On GstVideoDecoder::{drain,flush}, we send null packet with
CUVID_PKT_ENDOFSTREAM flag to drain out decoder. Which will
reset CUVID parser as well.
To continue decoding after the drain, the next input buffer
should include sequence headers otherwise CUVID parser will
not report any decodeable frame.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1911>
There could be a case where the new program has the same program number as the
previous one ... but is actually located on a PID previously used for elementary
stream. In that case the program is guaranteed to not be an update of the
previous program but a completely new one.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1893>
We need to be able to look for programs by their PID also. Using a hash table
was a bit sub-par (and overkill) for storing a range of programs.
This is needed because there could potentially be two programs with the same
program id but different PMT PID (while one is being deactivated the new one
would "exist").
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1893>
This commit modifies the interleave calculation to allow growing when incoming
data is before the segment start.
The rationale is that there is no requirement whatsoever for data before the
segment start to be "coherent" on all streams.
For example, a demuxer could rightfully send data from the video stream from the
previous keyframe (potentially quite a bit before the segment start) and the
audio from just before the segment start.
This will activate the same logic as growing the interleave when some streams
haven't received buffers yet.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1892>
* When a stream receives EOS, it will no longer change, we shouldn't take that
stream into account for interleave calculation.
* When streams (re)appear, we do not want to grow the initial interleave values
to excessive values. Instead of setting it to a default of 5s, progressively
grow it to that maximum.
* When the status of input streams change (i.e. going to/from "some haven't
received data yet" and "all have received data"), update the interleave
immediately instead of waiting for (potentially) 5s of data before updating
it.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1892>
The point here is that rtpsession will create a new rtpsource when
the field "rtx-ssrc" is present, and when not doing rtx, that means
a random ssrc will create a new rtpsource that will be included in RTCP
messages for the current session.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1882>
Instead of using GstMiniObject to hold H264 frame, now it uses a plain
structure. Besides, instead of holding a reference to
GstVideoCodecFrame, the H264 frame structure is set as a
GstVideoCodecFrame user data.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1856>
According to va_dec_hevc.h, pic_param->st_rps_bits should be set
for accelorater to skip parsing the *short_term_ref_pic_set
(num_short_term_ref_pic_sets) structure.
Also modified fill_picture to get parser info as a parameter,
in order to get slide_hdr->short_term_ref_pic_set_size.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1886>
Fix a small race where a group can receive stream-start
and post a pending buffering message just as another
thread posts a different buffering message, causing them
to be received by the application out of order. In the
worst case, this leads the application receiving a
stale 99% buffering message and going back to buffering
right after the 100% buffering message.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1840>
Apparently GtkSharp expects each object has only one ToggleRef at any
time. Assigning element.Handle into Raw has a consequence that second
ToggleRef attempts to get created but fails on g_object_unref () that
breaks a GObject assertion:
toggle_refs_notify: assertion failed: (tstack.n_toggle_refs == 1)
This is because toggle references should be removed with
g_object_remove_toggle_ref(), not a simple unref().
In order to avoid duplicate toggle references, introduce
ElementFactory.MakeRaw(), which creates a GstElement without its
accompanying C# object. The returned raw pointer can be assigned into
another GLib.Object without trouble.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1885>
And use the output segment position for the outgoing timestamp while it
is. This is needed to delay the calculation of `output_ts_offset` until
we actually have a usable timestamp, as tsmux will output a few initial
packets while `last_ts` is still unset.
Without this, the calculation would use the initial `0` value, which did
not have the intended effect of making VBR mode behave like CBR mode,
but always calculated an offset equal to the selected start time.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1884>
Fix suppression to support release and debug builds.
Here is the debug build call stack:
```
==10707== by 0x48B5520: g_malloc (gmem.c:106)
==10707== by 0x48D19DC: g_slice_alloc (gslice.c:1069)
==10707== by 0x48D3947: g_slist_copy_deep (gslist.c:619)
==10707== by 0x48D38B8: g_slist_copy (gslist.c:567)
==10707== by 0x4ADC90B: gst_debug_remove_with_compare_func (gstinfo.c:1504)
```
In release build `g_slist_copy (gslist.c:567)` got inlined:
```
==15419== by 0x48963E0: g_malloc (gmem.c:106)
==15419== by 0x48AA382: g_slice_alloc (gslice.c:1069)
==15419== by 0x48AB732: g_slist_copy_deep (gslist.c:619)
==15419== by 0x4A39B8F: gst_debug_remove_with_compare_func (gstinfo.c:1504)
```
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1814>
When doing only a single stream of audio/video this hardly matters,
but when doing many at the same time, the fact that you have to get
a hold of the glib global type-system lock every time you process a buffer,
means that there is a limit to how many streams you can process in
parallel.
Luckily the fix is very simple, by doing a cast rather than a full
type-check.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1873>
There is a chance that pool->buffers[index] sets BUFFER_STATE_QUEUED, but
it has not been queued yet which makes pool->buffers[index] still NULL.
At this time, if pool_streamff release all buffers with BUFFER_STATE_QUEUED
state regardless of whether the buffer is NULL or not, it will cause segfault.
To fix this, also check buffer when streamoff release buffer.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1842>
When tunneling over HTTP, if connection on the second channel happens
before the control timer is created we may trigger an assert in
rtsp_ctrl_timeout_remove(). Avoid that by taking the priv->lock before
attaching the client thread to the context.
Fixes#1025
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1867>
When there is vpp scaling downstream, we need to make sure SFC is not
triggered because vpp may fall into passthrough mode which causes
the decoder negotiation to create src caps with vpp scaled width/height.
This patch includes bitstream's original size in first query with
downstream in gst_msdkdec_src_caps, which is the same for what we do for
color format in this query. This is to ensure SFC scaling starts to
work only when downstream directly asks for a different size instead of
through vpp.
Note that here SFC scaling follows the same behavior as msdkvpp:
if user only changes width or height, e.g. dec ! video/x-raw,width=xx !,
the height will be modified to the value which fits the original DAR.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1838>
* Hide GstCudaMemory member variables
* Make GstCudaAllocator object GstCudaContext independent
* Set offset/stride of memory correctly via video meta
* Drop GST_BUFFER_POOL_OPTION_VIDEO_ALIGNMENT support.
This implementation actually does not support custom alignment
because we allocate device memory via cuMemAllocPitch
of which alignment is almost uncontrollable
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1834>
* cudaupload/download
- Specify only formats actually we can deal with
nvcodec elements, not all video formats
- Supports CUDA output for download and input for upload in order
to make passthrough possible, like other upload/download elements.
* cudabasetransform
- Reset conversion element if upstream CUDA memory
holds different CUDA context and the element can accept it.
This is the same behavior as corresponding d3d11 filter elements.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1834>