GitLab

The print_ref_pic_list_b now not only needs to trace the ref_pic_list_b0/1,
but also need to trace the ref_frame_list_0_short_term. We need to pass the
name directly to it rather than an index to refer to ref_pic_list_b0/1.

Part-of: <!2425>

The array sort of ref_frame_list_0_short_term has some typo. The
typo makes this list not in the POC ascend order and generate wrong
decoding result for interlaced streams.

Part-of: <!2425>

When driver return error on update plane request, kmssink
disables the scaling and retries plane update.
While doing so kmssink was matching the source rectangle dimensions
to the target rectangle dimensions which were calculated
as per scaling but this is incorrect, instead what we want here is
that target rectangle dimensions should match the source rectangle
dimensions as scaling is disabled now and so we match result
rectangle dimensions with source rectangle dimensions.

While at it, also match the result rectangle coordinates for
horizontal and vertical offsets with source rectange coordinates,
as since there is no scaling being done so no recentering is
required.

Part-of: <!2415>

Some cut VP9 streams begin with a non key frame. The current code
just bail out the parse_process_frame() if not a key frame. Because
of this, we do not set the valid caps before we push the data of the
first frame(even this first frame will be discarded by the downstream
decoder because it is not a key frame).

The pipeline such as:
gst-launch-1.0 filesrc location=some.ivf ! ivfparse ! vp9parse !
  vavp9dec ! fakesink
will get a negotiation error and the pipeline can not continue. The
correct behaviour should be: the decoder discard the first frame and
continue to decode later frames successfully.

So, when the parse does not have valid stream info(should be the first
frame case), we should continue and report caps.

Part-of: <!2427>

pipewiresrc outputs audio buffers without a valid duration, so we need
to calculate it manually in that case.

Upstream issue: pipewire/pipewire#1438

Part-of: <!2419>

We may need to drop the slices such as RASL pictures with the NoRaslOutputFlag, so
the current picture of h265decoder may be freed. We should not assign the frame->
output_buffer too early until we really output it. Or, the later coming slices will
allocate another picture and trigger the assert of:
  gst_video_decoder_allocate_output_frame_with_params:
  assertion 'frame->output_buffer == NULL' failed

Part-of: <!2421>

Some programs specify a PCR PID but don't actually store any PCR values, or are
way too far apart.

In order to gracefully handle those situations, we will queue up to a certain
amount of pending buffers before deciding to give up on that PCR PID and not use
any (i.e. using DTS/PTS values as-is)

Fixes #1629

Part-of: <!2422>

In H265, the stream may have odd bit depth such as 9 or 11. And
the bit depth of luma and chroma may differ. For example, the
stream with luma depth of 8 and chroma depth of 9 should use the
10 bit rtformat as the decoded picture format.

Part-of: <!2420>

In low_latency mode, try to bump the picture as soon as possible
without the frames disorder:
1. We can directly output the continuous non-reference frame.
2. Consider max_num_reorder_frames, which is special useful for
   I-P mode.
3. Consider the leading pictures with negative POC.
4  Output small POC pictures when non-reference frame comes.
4. Output the POC increment<=2 pictures. This is not 100% safe,
   but in practice this condition can be used.

Part-of: <!2373>

The max_num_reorder_frames can be useful for bump check. We store it
in the DPB and no need for the decoder now.

Part-of: <!2373>

The picture->ref field will change from time to time according to decoder's
state and reference sliding window. We need another flag to record whether
the picture is a reference picture when it is created, and this can help
the bumping check.

Part-of: <!2373>

The current behavior is different from the SPEC. We should check
and bump the DPB or drain the DPB before we insert the current
picture into it. This may cause the output picture disorder.

Part-of: <!2373>

Accord to spec, we should not add the current picture into the DPB
when we check whether it needs to bump, so the checks of the IDR and
the "memory_management_control_operation equal to 5" are no needed.

And the spec also says that the DPB only needs to bump when there is
no empty frame buffer left(We handle the IDR cases in other places).
We need to follow that and the max_num_reorder_frames is useless.

We also minus 1 in has_empty_frame_buffer because the current frame
has not been added yet.

Part-of: <!2373>

Part-of: <!2373>

When current frame memory_management_control_operation equal to 5, that
means we need to drain the dpb and the current picture act as an IDR frame.
So it should have smaller poc than the later pictures to ensure the output
order.

Part-of: <!2373>

grmbl

Part-of: <!2371>

* If we have an index table for non-framed essence, we can handle it

* The demuxer has a state which indicates whether it will next fetch a KLV or
data contained *within* a KLV.

* The position on Essence Tracks always correspond to the next entry to fetch,
demuxer offset will be skipped accordingly whenever we switch between
partitions (in case of resyncs). A copy of the main clip/custom KLV for that
partition is kept to track the position within the essence of that partition.

* For clip/custom-wrapped raw audio, if the edit rate is too small (and would
cause plenty of tiny buffers to be outputted), specify a minimum number of edit
units per buffer.

Part-of: <!2371>

* For pull-based, this avoids pulling content if it's not needed (ex: skipping filler
packet, not downloading the content if we only need to know if/where an essence
packet is, etc...). Allows reducing i/o usage to the minimum.

* This also allows doing sub-klv position tracking, and opens the way for
non-frame-wrapping handling

Part-of: <!2371>

This provides a summary of the number/type of tracks in the Material and File
Packages

Part-of: <!2371>

In order to figure out the exact start position (backed by a keyframe) accross
all tracks, we first figure out the backing keyframe position, and *then* seek
to that position.

Avoids ending up in situations where we would properly seek to the backing
keyframe on video ... but not on the audio streams (they would have been set to
the original non-keyframe position). Fixes key-unit seeking.

Part-of: <!2371>

Defined by Amendment 2:2013 to SMPTE ST 382:2007

Also define a new "UNKNOWN" wrapping type to make the difference with known
wrapping types

Part-of: <gstreamer/gst-plugins-bad!2371>

The picture essence coding matching was wrong. Use the proper "base" MXFUL for
video mpeg compression for matching.

Also handle the case where some old files would put the essence container label
in the essence coding field

Part-of: <gstreamer/gst-plugins-bad!2371>

* Streamline offset <=> entry handling. Historically the demuxer didn't support
information from index tables and stored the discovered information in an array
per track. When index table support was added, a parallel system was setup for
that relationship. This commit unifies this into one system with the
`find_edit_entry()` and `find_entry_for_offset()` functions.

* By extension, per-track offset entry tables are only created/used if no index
table is present for those tracks.

* Use index table information as-is. The index table system from MXF is quite
complex and there are various ways to use the information contained
within. Instead of converting that information we store the data from the tables
as-is and extract the needed information when needed.

* Handle index tables without entries (i.e. all content package units are of the
same size).

* Allow collecting index table segments as we go instead of only once if a
random-index-pack is present. This also improves support of some files in
push-mode.

* When searching for keyframe entries, use the keyframe_offset if
present (speeds up searching).

* For interleaved content (i.e. several tracks in the sample essence container),
we use a system to be able to identify the position of each track in the delta
entries of index tables.

* Handle temporal offset only on tracks which *do* need it (as specified in the
delta entries of the index tables). If present, those offsets are stored in a
pre-processed table which allows computing PTS from DTS with a simple offset.

* Add a quirk for files which are known to be have wrongly stored temporal
offsets.

* Overall opens the way to handle more types of MXF files, especially those with
non-frame-wrapping.

Part-of: <gstreamer/gst-plugins-bad!2371>

Part-of: <gstreamer/gst-plugins-bad!2371>

By printing out the various known flag values

Part-of: <gstreamer/gst-plugins-bad!2371>

A Source Clip can have zero'd SourcePackageID and SourceTrackID, this indicates
it terminates the source reference chain

Part-of: <gstreamer/gst-plugins-bad!2371>

This is similar to how the same issue was handled in qtdemux.

In order for the "DTS <= PTS" constraint to be respected, we calculate the
maximum temporal reordering that can happen (via index tables).

If there is a non-0 temporal reordering, we:
* Shift all outgoing PTS by that amount
* Shift segment for that stream by that amount
* Don't modify DTS (i.e. they might end up having negative running-time, before
the start of the segment)

Also ensure all entries have a valid PTS set, previously this wouldn't be set
for entries with a temporal offset of 0.

Fixes gstreamer/gst-plugins-bad#584

(and maybe a lot of other issues)

Part-of: <gstreamer/gst-plugins-bad!2371>

Added a comment with a future to-do, enhanced another comment and
fixed a typo in an error log message.

Part-of: <gstreamer/gst-plugins-bad!2417>

Move up gst_va_decoder_get_config() to group decoders function in the
same file area.

Part-of: <gstreamer/gst-plugins-bad!2417>

Change gst_va_decoder_format_changed() to
gst_va_decoder_config_is_equal(), which is more similar with other
GStreamer API.

The function call is replaced but it has to be negated because the
return value is the opposite.

Part-of: <gstreamer/gst-plugins-bad!2417>

Rename gst_va_decoder_change_resolution() to
gst_va_decoder_update_frame_size() which resembles
gst_va_decoder_set_frame_size().

Also added a comment to clarify the function use and makes more
specific the error message.

Part-of: <gstreamer/gst-plugins-bad!2417>

Renamed gst_va_decoder_set_format() to
gst_va_decoder_set_frame_size_with_surfaces() which resembles better
the passed parameters. Internally it creates the vaContext.

Added gst_va_decoder_set_frame_size() which is an alias of
gst_va_decoder_set_frame_size_with_surfaces() without surfaces. This
is the function which replaces gst_va_decoder_set_format() where
used.

Part-of: <gstreamer/gst-plugins-bad!2417>

We should use the NumPocTotalCurr value stored in decoder, which is a calculated
valid value, rather than use the invalid value in the slice header. Most of the
time, the NumPocTotalCurr is 0 and make the tmp_refs a very short length, and
causes the decoder's wrong result.
By the way, the NumPocTotalCurr is not the correct name specified in H265 spec,
its name should be NumPicTotalCurr. We change it to the correct name.

Part-of: <gstreamer/gst-plugins-bad!2414>

The VA's ReferenceFrames should only contain the reference frame, we
should not add the non reference frames into this list.

Part-of: <gstreamer/gst-plugins-bad!2414>

... if closedcaption plugin is available

Part-of: <!2411>

Some GPUs support BGRA format and it will be converted to subsampled
YUV format by GPU internally. Disable this implicit conversion
since the conversion parameters such as input/output colorimetry
are not exposed nor it's written in bitstream (e.g., VUI).
We prefer explicit conversion via our conversion elements.

Part-of: <gstreamer/gst-plugins-bad!2410>

The temp guint8 var of delta_chroma_offset_l0 and delta_chroma_offset_l1
can not cover the full range of delta_chroma_weight_l0/1 in the slice
header. When overflow happens, the decoder result is wrong.

Part-of: <gstreamer/gst-plugins-bad!2412>

Some manifests use the ContentProtection node to store additional information
such as the license server url. Our MPD parser used to process the
ContentProtection node, extracting Playready PSSH boxes. However for other DRM
systems, only the `value` attribute was passed down to the protection event, so
for example, Widevine data was not parsed at all and "Widevine" was passed to
the event, which is not very useful for decryptors that require a PSSH init
data.

Parsing should now be done by decryptors which will receive the entire
ContentProtection XML node as a string. This gives more "freedom" to the
decryptor which can then detect and parse custom nodes as well.

Part-of: <!2400>

The VP9 streams have the ability to change the resolution dynamically
at any time point. It does not send ad KEY frame before change the
resolution, even the INTER frame can change the resolution immediately.
So we need to check the resolution change for each frame and do the
re-negiotiation if needed.

Some insaned stream may play in resolution A first and then dynamically
changes to B, and after 1 or 2 frames, it use a show_existing_frame to
repeat the old frame of resolution A before. So, not only new_picture(),
but also duplicate_picture() need to check this.

Part-of: <gstreamer/gst-plugins-bad!2407>

The driver for VP9 should have the ability to handle the dynamical resolution
changes. So if only the resolution changes, we should not re-create the config
and context in negotiation.

Part-of: <gstreamer/gst-plugins-bad!2407>