platform/upstream/gstreamer.git
3 years agosrtsink: remove unused connection_mode variable
Jakub Adam [Tue, 25 Aug 2020 11:44:42 +0000 (13:44 +0200)]
srtsink: remove unused connection_mode variable

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1541>

3 years agosrtobject: obey "wait-for-connection" in caller mode
Jakub Adam [Mon, 23 Nov 2020 15:12:39 +0000 (16:12 +0100)]
srtobject: obey "wait-for-connection" in caller mode

The pipeline now gets stuck in gst_srt_object_write_one() until the
receiver comes online, which may or may not be desired based on the use
case.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1836>

3 years agova: mpeg2dec: refactor the picture reference filling
Víctor Manuel Jáquez Leal [Tue, 5 Jan 2021 13:18:39 +0000 (14:18 +0100)]
va: mpeg2dec: refactor the picture reference filling

Add the helper function _get_surface_id() which extracts the
VASurfaceID from the passed picture. This function gets the surface of
the next and previous reference picture.

Instead of if-statements, this refactor uses a switch-statement with a
fall-through, for P-type pictures, making the code a bit more readable.

Also it adds quirks for gallium driver, which cannot handle invalid
surfaces as forwarding nor backwarding references, so the function fails.
Also iHD cannot handle them, but to avoid failing, the current picture
is used as self-reference.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1939>

3 years agova: mpeg2dec: set first field either frame or has a first field
Víctor Manuel Jáquez Leal [Tue, 5 Jan 2021 13:16:45 +0000 (14:16 +0100)]
va: mpeg2dec: set first field either frame or has a first field

Add a helper function _is_frame_start() which check if picture has a
frame structure or if it has not an interlaced first field yet. This
function is used with filling is_first_field parameter.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1939>

3 years agocodecs: mpeg2decoder: decode only if B and not closed gop
Víctor Manuel Jáquez Leal [Wed, 6 Jan 2021 15:38:14 +0000 (16:38 +0100)]
codecs: mpeg2decoder: decode only if B and not closed gop

Mark as decode only if picture type is B, without previous picture in DBP and
closed_gop is 0 as might be understood in "6.3.8 Group of pictures header".

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1939>

3 years agocodecs: mpeg2decoder: rename variables
Víctor Manuel Jáquez Leal [Wed, 6 Jan 2021 11:48:14 +0000 (12:48 +0100)]
codecs: mpeg2decoder: rename variables

Since prev_picture and next_picture are plain pointers, not pointer to pointers,
it's misleading to name them with _ptr suffix.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1939>

3 years agova: display: parse and set driver implementation
Víctor Manuel Jáquez Leal [Mon, 4 Jan 2021 20:02:35 +0000 (21:02 +0100)]
va: display: parse and set driver implementation

This enum can be used for quirk handling. It's not a property because
the driver enum list might change, it's not static, thus avoiding the
update of GType declaration.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1938>

3 years agova: display: add function precondition check
Víctor Manuel Jáquez Leal [Mon, 4 Jan 2021 19:56:26 +0000 (20:56 +0100)]
va: display: add function precondition check

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1938>

3 years agosrtobject: post a message on the bus when broken socket is detected
Jakub Adam [Tue, 25 Aug 2020 17:12:13 +0000 (19:12 +0200)]
srtobject: post a message on the bus when broken socket is detected

So that the application gets notified may react to it.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1935>

3 years agova: mpeg2dec: Using the current picture's surface when missing reference.
He Junyan [Wed, 30 Dec 2020 15:29:47 +0000 (23:29 +0800)]
va: mpeg2dec: Using the current picture's surface when missing reference.

When missing the reference frames, we should not just discard the current
frame. Some streams have group of picture header. It is an optional header
that can be used immediately before a coded I-frame to indicate to the decoder
if the first consecutive B-pictures immediately following the coded I-frame can
be reconstructed properly in the case of a random access.
In that case, the B frames may miss the previous reference and can still be
correctly decoded. We also notice that the second field of the I frame may
be set to P type, and it only ref its first field.
We should not skip all those frames, and even the frame really misses the
reference frame, some manner such as inserting grey picture should be used
to handle these cases.

The driver crashes when it needs to access the reference picture while we set
forward_reference_picture or backward_reference_picture to VA_INVALID_ID. We
now set it to current picture to avoid this. This is just a temp manner.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1929>

3 years agocodecs: mpeg2decoder: Creating the field based on its arriving time.
He Junyan [Wed, 30 Dec 2020 15:14:01 +0000 (23:14 +0800)]
codecs: mpeg2decoder: Creating the field based on its arriving time.

Spec says:
In a frame picture top_field_first being set to ‘1’ indicates that the
top field of the reconstructed frame is the first field output by the
decoding process. top_field_first being set to ‘0’ indicates that the
bottom field of the reconstructed frame is the first field output by
decoding process.

Here, the "output" should be interpreted just as the output order, not
including the decoding order. The field should be decoded as the order
they comes in the stream. Namely, no matter top_field_first is 0 or 1,
the first coming field is the first one to be decoded.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1929>

3 years agova: mpeg2dec: Apply buffer_flags to the output buffer.
He Junyan [Fri, 1 Jan 2021 08:00:10 +0000 (16:00 +0800)]
va: mpeg2dec: Apply buffer_flags to the output buffer.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1929>

3 years agocodecs: Add buffer_flags for mpeg2 picture.
He Junyan [Fri, 1 Jan 2021 07:56:03 +0000 (15:56 +0800)]
codecs: Add buffer_flags for mpeg2 picture.

We need to store the buffer flags such as GST_VIDEO_BUFFER_FLAG_INTERLACED
and GST_VIDEO_BUFFER_FLAG_TFF for interlaced video. Without these flags,
the VPP and display elements can not apply filter correctly.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1929>

3 years agocodecs: Reset the quant matrices for each sequence in mpeg2 decoder.
He Junyan [Wed, 30 Dec 2020 15:00:51 +0000 (23:00 +0800)]
codecs: Reset the quant matrices for each sequence in mpeg2 decoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1929>

3 years agosrt: Add authentication to srtsink and srtsrc elements
Raghavendra [Tue, 27 Oct 2020 06:22:09 +0000 (11:52 +0530)]
srt: Add authentication to srtsink and srtsrc elements

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1725>

3 years agodashsink: fix critical log when exit dynamic pipeline
Haihua Hu [Wed, 30 Dec 2020 14:52:01 +0000 (22:52 +0800)]
dashsink: fix critical log when exit dynamic pipeline

availability-start-time and publish-time shared the same
GstDateTime object, this object will be unref twice and
cause reference count issue. Should use g_value_dup_boxed()
to copy this object.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1928>

3 years agomsdkenc{h264,h265}: add min-qp and max-qp properties
Haihao Xiang [Wed, 23 Dec 2020 08:11:42 +0000 (16:11 +0800)]
msdkenc{h264,h265}: add min-qp and max-qp properties

The SDK allows user to set a QP range [1], so add min-qp and max-qp to
sepecify QP range. By default, there is no limitations on QP.

[1] https://github.com/Intel-Media-SDK/MediaSDK/blob/master/doc/mediasdk-man.md#mfxextcodingoption2

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1908>

3 years agomsdkenc{h264,h265}: add p-pyramid property
Haihao Xiang [Wed, 23 Dec 2020 05:36:02 +0000 (13:36 +0800)]
msdkenc{h264,h265}: add p-pyramid property

The SDK can support P-Pyramid reference structure [1], so add a new
property to enable this feature in msdkenc{h264,h265}.

[1] https://github.com/Intel-Media-SDK/MediaSDK/blob/master/doc/mediasdk-man.md#preftype

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1908>

3 years agomsdkh265enc: add b-pyramid property
Haihao Xiang [Tue, 22 Dec 2020 08:17:18 +0000 (16:17 +0800)]
msdkh265enc: add b-pyramid property

Like as msdkh264enc, b-pyramid is added to enable B-Pyramid reference
structure for H265 encoding

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1908>

3 years agomsdkh265enc: add transform-skip property
Haihao Xiang [Tue, 22 Dec 2020 06:54:59 +0000 (14:54 +0800)]
msdkh265enc: add transform-skip property

Since the SDK API 1.26, TransformSkip was added to control
transform_skip_enabled_flag setting in PPS [1]

[1] https://github.com/Intel-Media-SDK/MediaSDK/blob/master/doc/mediasdk-man.md#mfxextcodingoption3

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1908>

3 years agomsdk: needn't close mfx session when failed
Haihao Xiang [Tue, 29 Dec 2020 01:41:05 +0000 (09:41 +0800)]
msdk: needn't close mfx session when failed

Otherwise we will get double free issue because mfx session is closed in
finalize. See
https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1867#note_739346
for the double free issue.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1916>

3 years agowebrtcstats: PLI/FIR/NACK direction are the opposite of the media
Olivier Crête [Tue, 29 Dec 2020 18:29:05 +0000 (13:29 -0500)]
webrtcstats: PLI/FIR/NACK direction are the opposite of the media

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1924>

3 years agoassrender: Don't try unlocking unlocked mutex
Sebastian Dröge [Tue, 29 Dec 2020 11:15:10 +0000 (13:15 +0200)]
assrender: Don't try unlocking unlocked mutex

When flushing right at the beginning of the video chain function or
when failing negotiation at the top of the function, the assrender mutex
would be unlocked without being previously locked.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1918>

3 years agod3d11compositor: Add support for resolution change
Seungha Yang [Sun, 27 Dec 2020 13:16:13 +0000 (22:16 +0900)]
d3d11compositor: Add support for resolution change

Not only for position update (e.g., xpos, ypos),
we need to configure shader again per resolution change of each
input stream.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1912>

3 years agod3d11shader: Fix ID3DBlob object leak
Seungha Yang [Sun, 27 Dec 2020 19:33:11 +0000 (04:33 +0900)]
d3d11shader: Fix ID3DBlob object leak

Even if HLSL compiler was able to compile our shader code, D3DCompile()
might return ID3DBlob object for compile warnings and the object
should be released.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1914>

3 years agocodecs: Fix a typo in mpeg2 stateless decoder base class.
He Junyan [Mon, 28 Dec 2020 09:13:22 +0000 (17:13 +0800)]
codecs: Fix a typo in mpeg2 stateless decoder base class.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1915>

3 years agova: mpeg2dec: cosmetic changes
Víctor Manuel Jáquez Leal [Thu, 24 Dec 2020 19:07:09 +0000 (20:07 +0100)]
va: mpeg2dec: cosmetic changes

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1798>

3 years agocodecs: mpeg2decoder: fix documentation
Víctor Manuel Jáquez Leal [Sun, 27 Dec 2020 14:47:13 +0000 (15:47 +0100)]
codecs: mpeg2decoder: fix documentation

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1798>

3 years agocodecs: mpeg2decoder: simplify macros
Víctor Manuel Jáquez Leal [Thu, 24 Dec 2020 15:20:31 +0000 (16:20 +0100)]
codecs: mpeg2decoder: simplify macros

For constructors, instead of casting to pointers, cast to the structures.

For compare, use inlined functions.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1798>

3 years agova: Add mpeg2 VA decoder.
He Junyan [Fri, 18 Dec 2020 14:28:41 +0000 (22:28 +0800)]
va: Add mpeg2 VA decoder.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1798>

3 years agocodecs: Add mpeg2 stateless decoder base class.
He Junyan [Fri, 18 Dec 2020 13:25:08 +0000 (21:25 +0800)]
codecs: Add mpeg2 stateless decoder base class.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1798>

3 years agomfvideoenc: Re-define default GOP size value
Seungha Yang [Sat, 26 Dec 2020 18:16:28 +0000 (03:16 +0900)]
mfvideoenc: Re-define default GOP size value

The behavior for zero AVEncMPVGOPSize value would be
varying depending on GPU vendor implementation and some
GPU will produce keyframe only once at the beginning of encoding.
That's unlikely expected result for users.

To make this property behave consistently among various GPUs,
this commit will change default value of "gop-size" property to -1
which means "auto". When "gop-size" is unspecified, then
mfvideoenc will calculate GOP size based on framerate
like that of our x264enc implementation.

See also
https://docs.microsoft.com/en-us/windows/win32/directshow/avencmpvgopsize-property

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1911>

3 years agomfvideoenc: Fix use of uninitialized value
Seungha Yang [Sat, 26 Dec 2020 18:43:11 +0000 (03:43 +0900)]
mfvideoenc: Fix use of uninitialized value

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1911>

3 years agod3d11device: Add property for getting adapter LUID
Seungha Yang [Thu, 24 Dec 2020 12:31:04 +0000 (21:31 +0900)]
d3d11device: Add property for getting adapter LUID

LUID (Locally Unique Identifier) can used for identifying GPU
and that's required for some Windows APIs (e.g., MFTEnum2()) to setup device.

See also
https://docs.microsoft.com/en-us/windows/win32/api/mfapi/nf-mfapi-mftenum2

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1910>

3 years agomfvideoenc: Remove duplicated class registration code
Seungha Yang [Sat, 26 Dec 2020 11:39:07 +0000 (20:39 +0900)]
mfvideoenc: Remove duplicated class registration code

Each codec subclass has the same code for class/element registration,
so we can move the code into one helper methodm and that will make
future enhancement simple.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1909>

3 years agomsdk: check GstMsdkContext instead of mfxSession instance
Haihao Xiang [Thu, 10 Dec 2020 03:11:04 +0000 (11:11 +0800)]
msdk: check GstMsdkContext instead of mfxSession instance

When creating a GstMsdkContext instance, it also creates a mfxSession
instance, so we may check GstMsdkContext instead of mfxSession instance
to make sure MSDK is available. In addition, according to MSDK doc [1],
MFXVideoCORE_SetHandle function should be executed before any actual
usage of library including queries, otherwise the behavior is
unexpected, so we should call MFXVideoCORE_QueryPlatform after
MFXVideoCORE_SetHandle on Linux

[1] https://github.com/Intel-Media-SDK/MediaSDK/blob/master/doc/mediasdk-man.md#working-with-va-api-applications

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1867>

3 years agod3d11: Remove unnecessary helper methods
Seungha Yang [Wed, 23 Dec 2020 12:21:55 +0000 (21:21 +0900)]
d3d11: Remove unnecessary helper methods

We can query selected D3D_FEATURE_LEVEL and factory version
by using native D3D11 API

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1906>

3 years agod3d11: Re-implement Desktop Duplication source
Seungha Yang [Fri, 20 Nov 2020 18:20:36 +0000 (03:20 +0900)]
d3d11: Re-implement Desktop Duplication source

Add a new video source element "d3d11desktopdupsrc" for capturing desktop image
via Desktop Duplication based on Microsoft's Desktop Duplication sample available at
https://github.com/microsoft/Windows-classic-samples/tree/master/Samples/DXGIDesktopDuplication

This element is expected to be a replacement of existing dxgiscreencapsrc
element in winscreencap plugin.

Currently this element can support (but dxgiscreencapsrc cannot)
- Copying captured D3D11 texture to output buffer without download
- Support desktop session transition
  e.g., can capture desktop without error even in case that
  "Lock desktop" and "Permission dialog"
- Multiple d3d11desktopdupsrc elements can capture the same monitor

Not yet implemented features
- Cropping rect is not implemented, but that can be handled by downstream
- Mult-monitor is not supported. But that is also can be implemented by
  downstream element for example via multiple d3d11desktopdup elements
  with d3d11compositor

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1855>

3 years agod3d11device: Add an optional flags argument for creating device
Seungha Yang [Mon, 21 Dec 2020 15:47:09 +0000 (00:47 +0900)]
d3d11device: Add an optional flags argument for creating device

Extend gst_d3d11_device_new() method so that caller can specify
D3D11_CREATE_DEVICE_FLAG value to use.

See https://docs.microsoft.com/en-us/windows/win32/api/d3d11/ne-d3d11-d3d11_create_device_flag
for more detail about D3D11_CREATE_DEVICE_FLAG

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1901>

3 years agodvbsuboverlay: Add support for dynamic resolution update.
Raju Babannavar [Mon, 21 Dec 2020 08:36:53 +0000 (14:06 +0530)]
dvbsuboverlay: Add support for dynamic resolution update.

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1487

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1897>

3 years agod3d11device: Remove dead code
Seungha Yang [Sun, 20 Dec 2020 17:56:55 +0000 (02:56 +0900)]
d3d11device: Remove dead code

We don't use this method since the commit of
0788492461e1b559230cc5c3a354fe5f48f95f8b

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1892>

3 years agod3d11: Privatize d3d11memory implementation
Seungha Yang [Sat, 19 Dec 2020 17:39:40 +0000 (02:39 +0900)]
d3d11: Privatize d3d11memory implementation

Hide most of symbols of GstD3D11Memory object.
GstD3D11Memory is one of primary resource for imcoming d3d11 library
and it's expected to be a extensible feature.
Hiding implementation detail would be helpful for later use case.

Summary of this commit:
* Now all native Direct3D11 resources are private of GstD3D11Memory.
  To access native resources, getter methods need to be used
  or generic map (e.g., gst_memory_map) API should be called
  apart from some exceptional case such as d3d11decoder case.
* Various helper methods are added for GstBuffer related operations
  and in order to remove duplicated code.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1892>

3 years agod3d11: Add a helper method for d3d11buffferpool setup
Seungha Yang [Sat, 19 Dec 2020 16:06:24 +0000 (01:06 +0900)]
d3d11: Add a helper method for d3d11buffferpool setup

Remove duplicated code for d3d11buffferpool setup.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1892>

3 years agod3d11device: Remove optional helper methods
Seungha Yang [Fri, 18 Dec 2020 15:40:53 +0000 (00:40 +0900)]
d3d11device: Remove optional helper methods

Most of Direct3D11 APIs can be called without GstD3D11Device
abstraction. This is a part of prework for public GstD3D11 library
to introduce minimal APIs

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1892>

3 years agod3d11videosink: Prepare window once streaming started
Seungha Yang [Sun, 20 Dec 2020 13:12:44 +0000 (22:12 +0900)]
d3d11videosink: Prepare window once streaming started

... instead of READY state. READY state is too early for setting
overlay window handle especially playbin/playsink scenario
since playsink will set given overlay handle on videosink once
READY state change of videosink is ensured.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1893>

3 years agomfvideoenc: Improve latency performance for hardware encoder
Seungha Yang [Tue, 18 Aug 2020 18:19:26 +0000 (03:19 +0900)]
mfvideoenc: Improve latency performance for hardware encoder

Unlike software MFT (Media Foundation Transform) which is synchronous
in terms of processing input and output data, hardware MFT works
in asynchronous mode. output data might not be available right after
we pushed one input data into MFT.
Note that async MFT will fire two events, one is "METransformNeedInput"
which happens when MFT can accept more input data,
and the other is "METransformHaveOutput", that's for signaling
there's pending data which can be outputted immediately.

To listen the events, we can wait synchronously via
IMFMediaEventGenerator::GetEvent() or make use of IMFAsyncCallback
object which is asynchronous way and the event will be notified
from Media Foundation's internal worker queue thread.

To handle such asynchronous operation, previous working flow was
as follows (IMFMediaEventGenerator::GetEvent() was used for now)
- Check if there is pending output data and push the data toward downstream.
- Pulling events (from streaming thread) until there's at least
  one pending "METransformNeedInput" event
- Then, push one data into MFT from streaming thread
- Check if there is pending "METransformHaveOutput" again.
  If there is, push new output data to downstream
  (unlikely there is pending output data at this moment)

Above flow was processed from upstream streaming thread. That means
even if there's available output data, it could be outputted later
when the next buffer is pushed from upstream streaming thread.
It would introduce at least one frame latency in case of live stream.

To reduce such latency, this commit modifies the flow to be fully
asynchronous like hardware MFT was designed and to be able to
output encoded data whenever it's available. More specifically,
IMFAsyncCallback object will be used for handling
"METransformNeedInput" and "METransformHaveOutput" events from
Media Foundation's internal thread, and new output data will be
also outputted from the Media Foundation's thread.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1520>

3 years agodecklinkaudiosrc: Fix duration of the first audio frame after each discont
Sebastian Dröge [Wed, 16 Dec 2020 16:32:25 +0000 (18:32 +0200)]
decklinkaudiosrc: Fix duration of the first audio frame after each discont

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1886>

3 years agomediafoundation: Fix redefinition of variables.
Biswapriyo Nath [Tue, 15 Dec 2020 18:58:08 +0000 (00:28 +0530)]
mediafoundation: Fix redefinition of variables.

Remove duplicate GstMFDevice and GstMFDeviceProvider declaration.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1884>

3 years agoaudiobuffersplit: Calculate the correct size for fixed size buffers
Jan Schmidt [Wed, 16 Dec 2020 17:41:18 +0000 (04:41 +1100)]
audiobuffersplit: Calculate the correct size for fixed size buffers

Fix the output-buffer-size property to do what it says by calculating
the correct audio buffer size for that target size, rounded down to
the nearest whole number of samples.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1887>

3 years agodecklink: Implement GstBaseSrc::get_caps() to return more constrained caps
Sebastian Dröge [Thu, 10 Dec 2020 10:35:07 +0000 (12:35 +0200)]
decklink: Implement GstBaseSrc::get_caps() to return more constrained caps

Instead of the template caps we can return a subset of them based on the
selected properties.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1868>

3 years agowasapi2: Ensure unmute when opening audio client
Seungha Yang [Thu, 29 Oct 2020 17:21:11 +0000 (02:21 +0900)]
wasapi2: Ensure unmute when opening audio client

ISimpleAudioVolume::SetMute() status seems to be preserved even
after process is terminated. In order to start audio client with
unmuted state, always disable mute when opening audio client.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1731>

3 years agotsparse: Don't use non-object for debugging statement
Edward Hervey [Mon, 14 Dec 2020 15:12:22 +0000 (16:12 +0100)]
tsparse: Don't use non-object for debugging statement

Use the pad instead

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1880>

3 years agoexamples/ts-parser: Use the section type for descriptor identification
Edward Hervey [Mon, 14 Dec 2020 09:56:39 +0000 (10:56 +0100)]
examples/ts-parser: Use the section type for descriptor identification

Some descriptors can only be present in some section

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1880>

3 years agoexamples/ts-parser: Try more descriptor/stream types
Edward Hervey [Mon, 14 Dec 2020 09:56:02 +0000 (10:56 +0100)]
examples/ts-parser: Try more descriptor/stream types

These were added recently

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1880>

3 years agompegts: Don't add non-padded streams to collection on updates
Edward Hervey [Wed, 9 Dec 2020 08:14:12 +0000 (09:14 +0100)]
mpegts: Don't add non-padded streams to collection on updates

When carrying over existing GstStream to a new GstStreamCollection we need to
check whether they *actually* were being used in the previous collection.

This avoids adding unknown streams (metadata, PSI, etc...) to the collection on
updates.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1880>

3 years agompegts: Add support for SIT sections
Edward Hervey [Sun, 22 Nov 2020 17:48:08 +0000 (18:48 +0100)]
mpegts: Add support for SIT sections

Selection Information Tables (EN 300 468)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1852>

3 years agompegts: Update documentation
Edward Hervey [Mon, 14 Dec 2020 09:50:02 +0000 (10:50 +0100)]
mpegts: Update documentation

* Split up into appropriate individual header files
* Document more sections and structures
* Add well-known list of registration id

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1879>

3 years agoplayer/transcoder: Use bus signal watch
Thibault Saunier [Thu, 10 Dec 2020 19:29:31 +0000 (16:29 -0300)]
player/transcoder: Use bus signal watch

Instead of implementing exactly the same thing ourself but making
`GstBus` not know that it is the case.

Since we are *sure* that the bus can't have been access at the point
where we add the watch we are guaranteed that the current thread
maincontext is going to be used.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1870>

3 years agointervideosrc: fix negotiation of interlaced caps
Lim Siew Hoon [Thu, 10 Dec 2020 07:37:14 +0000 (15:37 +0800)]
intervideosrc: fix negotiation of interlaced caps

In 1.0 the field in caps is called "interlace-mode", not "interlaced".

Fixes #1480

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1869>

3 years agoopenaptx: Drop lib prefix from option name for consistency
Arun Raghavan [Sat, 12 Dec 2020 02:45:25 +0000 (21:45 -0500)]
openaptx: Drop lib prefix from option name for consistency

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1876>

3 years agoopenaptx: add aptX and aptX-HD codecs using libopenaptx
Igor Kovalenko [Fri, 11 Dec 2020 08:45:06 +0000 (08:45 +0000)]
openaptx: add aptX and aptX-HD codecs using libopenaptx

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1871>

3 years agowpe: Emit load-progress messages
Philippe Normand [Mon, 19 Oct 2020 13:56:43 +0000 (14:56 +0100)]
wpe: Emit load-progress messages

The estimated-load-progress value can be used on application side to display a
progress bar for instance.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1710>

3 years agobasetsmux: Don't send the capsheader if src pad has no caps
Vivia Nikolaidou [Tue, 8 Dec 2020 14:46:42 +0000 (16:46 +0200)]
basetsmux: Don't send the capsheader if src pad has no caps

That means we're shutting down, so there's no point in the streamheader
being sent

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1864>

3 years agortmp2/connection: pass the parent cancellable down to the connection
Matthew Waters [Fri, 4 Dec 2020 06:02:00 +0000 (17:02 +1100)]
rtmp2/connection: pass the parent cancellable down to the connection

Otherwise, when rtpm2src cancels an inflight operation that has a queued
message stored, then the rtmp connection operation is not stopped.

If the cancellation occurs during rtmp connection start up, then
rtpm2src does not have any way of accessing the connection object as it
has not been returned yet.  As a result, rtpm2src will cancel, the
connection will still be processing things and the
GMainContext/GMainLoop associated with the outstanding operation will be
destroyed.  All outstanding operations and the rtmpconnection object will
therefore be leaked in this case.

Fixes: https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1425
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1862>

3 years agosrt: Don't take object lock calling gst_srt_object_get_stats
Jan Alexander Steffens (heftig) [Mon, 7 Dec 2020 13:54:28 +0000 (14:54 +0100)]
srt: Don't take object lock calling gst_srt_object_get_stats

This function takes the sock lock. This can result in a deadlock when
another thread holding the sock lock is trying to take the object lock.

Thread A (Holds object lock, wants sock lock):

    #2  gst_srt_object_get_stats at gst-plugins-bad/ext/srt/gstsrtobject.c:1753
    #3  gst_srt_object_get_property_helper at gst-plugins-bad/ext/srt/gstsrtobject.c:409
    #4  gst_srt_sink_get_property at gst-plugins-bad/ext/srt/gstsrtsink.c:95
    #5  g_object_get_property from libgobject-2.0.so.0

Thread B (Holds sock lock, wants object lock):

    #2  gst_element_post_message_default at gstreamer/gst/gstelement.c:2069
    #3  gst_element_post_message at gstreamer/gst/gstelement.c:2123
    #4  gst_element_message_full_with_details at gstreamer/gst/gstelement.c:2259
    #5  gst_element_message_full at gstreamer/gst/gstelement.c:2298
    #6  gst_srt_object_send_headers at gst-plugins-bad/ext/srt/gstsrtobject.c:1407
    #7  gst_srt_object_send_headers at gst-plugins-bad/ext/srt/gstsrtobject.c:1444
    #8  gst_srt_object_write_to_callers at gst-plugins-bad/ext/srt/gstsrtobject.c:1444
    #9  gst_srt_object_write at gst-plugins-bad/ext/srt/gstsrtobject.c:1598
    #10 gst_srt_sink_render at gst-plugins-bad/ext/srt/gstsrtsink.c:179

Fixes d2d00e07acc2b1ab1ae5a728ef5dc33c9dee7869.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1861>

3 years agoccconverter: Add property to specify which sections to include in CDP packets
Sebastian Dröge [Wed, 25 Nov 2020 14:24:25 +0000 (16:24 +0200)]
ccconverter: Add property to specify which sections to include in CDP packets

Various software, including ffmpeg's Decklink support, fails parsing CDP
packets that contain anything but CC data in the CDP packets.

Based on this property, timecodes are not written into the CDP packets
even if they're present.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1833>

3 years agoccconverter: Refactor code to only retrieve the timecode meta once
Sebastian Dröge [Wed, 25 Nov 2020 12:54:09 +0000 (14:54 +0200)]
ccconverter: Refactor code to only retrieve the timecode meta once

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1833>

3 years agova: decode: fix display type
Víctor Manuel Jáquez Leal [Sun, 6 Dec 2020 17:03:47 +0000 (18:03 +0100)]
va: decode: fix display type

Instead of a pointer to GstVaDisplay it was used a VADisplay type, which in
certain platforms is the same, and the compiler didn't complain.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1860>

3 years agortpmanagerbad: allow setting caps on rtpsrc
Marc Leeman [Fri, 3 Jul 2020 10:25:31 +0000 (12:25 +0200)]
rtpmanagerbad: allow setting caps on rtpsrc

rtpsrc tries to do a lookup of the caps based on the encoding-name. For
not so standard encodings, the caps can be set, avoiding the lookup.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1406>

3 years agod3d11videosink: Add a property to support rendering statistics data on window
Seungha Yang [Sat, 21 Nov 2020 19:39:57 +0000 (04:39 +0900)]
d3d11videosink: Add a property to support rendering statistics data on window

Add a new property "render-stats" to allow rendering statistics
data on window for debugging and/or development purpose.
Text rendering will be accelerated by GPU since this implementation
uses Direct2D/DirectWrite API and Direct3D inter-op for minimal overhead.
Specifically, text data will be rendered on swapchain backbuffer
directly without any copy/allocation of extra texture.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1830>

3 years agod3d11: Protect ID3D11VideoContext with lock
Seungha Yang [Thu, 3 Dec 2020 18:40:17 +0000 (03:40 +0900)]
d3d11: Protect ID3D11VideoContext with lock

Likewise d3d11 immediate context (i.e., ID3D11DeviceContext),
ID3D11VideoContext API is not thread safe. It must be protected therefore.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1856>

3 years agodocs: don't exit the subdir when optional deps aren't found
Mathieu Duponchelle [Thu, 3 Dec 2020 16:13:15 +0000 (17:13 +0100)]
docs: don't exit the subdir when optional deps aren't found

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1854>

3 years agoopencv: Expose retinex parameters
Edward Hervey [Wed, 2 Dec 2020 10:29:08 +0000 (11:29 +0100)]
opencv: Expose retinex parameters

Makes the plugin a tad more useful :)

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1845>

3 years agogst-libs/gst/wayland: Install "unstable" wayland header
Marius Vlad [Mon, 12 Oct 2020 11:12:07 +0000 (14:12 +0300)]
gst-libs/gst/wayland: Install "unstable" wayland header

Context creation and retrieval is required, the symbols are exported
with the header missing. Users most likely define GST_USE_UNSTABLE_API
so they're aware of the implications of using a header that might change
between releases.

Signed-off-by: Marius Vlad <marius.vlad@collabora.com>
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1688>

3 years agohlsdemux: Use actual object for logging
Edward Hervey [Thu, 3 Dec 2020 13:12:06 +0000 (14:12 +0100)]
hlsdemux: Use actual object for logging

i.e. the pad of the stream

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1853>

3 years agocurl: Remove incorrect GST_DEBUG_OBJECT() calls
Arun Raghavan [Thu, 3 Dec 2020 11:55:00 +0000 (06:55 -0500)]
curl: Remove incorrect GST_DEBUG_OBJECT() calls

klass is not a GstObject, and these debugs print should likely not be
around anyway.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1851>

3 years agocuda: Fix lowest targetted architecture for CUDA >= 11.0
Edward Hervey [Wed, 25 Nov 2020 16:59:54 +0000 (17:59 +0100)]
cuda: Fix lowest targetted architecture for CUDA >= 11.0

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1469

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1835>

3 years agotsparse: Forward incoming timestamps
Edward Hervey [Thu, 5 Nov 2020 11:48:27 +0000 (13:48 +0200)]
tsparse: Forward incoming timestamps

Ensure we properly forward the upstream PTS/DTS on the regular and program
source pads. All packets being processed will carry over the latest PTS/DTS (as
a reconstructed GstBuffer).

Fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1419

And properly forward PTS/DTS for program pads (which wasn't the case before)

Original patch by Vivia Nikolaidou <vivia@ahiru.eu>

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1769>

3 years agoadaptivedemux: Don't log with non-GObject objects
Sebastian Dröge [Wed, 2 Dec 2020 07:39:45 +0000 (09:39 +0200)]
adaptivedemux: Don't log with non-GObject objects

Instead of using the streams, log with the pad of the streams.

https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/issues/1457

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1844>

3 years agotranscodebin: Minor error message enhancement
Thibault Saunier [Fri, 20 Nov 2020 14:29:46 +0000 (11:29 -0300)]
transcodebin: Minor error message enhancement

3 years agotranscodebin: Unlock while setting decodebin caps
Thibault Saunier [Fri, 20 Nov 2020 01:56:46 +0000 (22:56 -0300)]
transcodebin: Unlock while setting decodebin caps

Otherwise it will deadlock recursing up to notify parent object property changes

3 years agotranscodebin: Avoid plugin converter if filter handles ANY caps
Thibault Saunier [Thu, 19 Nov 2020 21:31:34 +0000 (18:31 -0300)]
transcodebin: Avoid plugin converter if filter handles ANY caps

For example identity or clocksync or this kind of elements can be
used with any data flow and we should not enforce decoding to row in
that case.

3 years agotranscodebin: Add filter as soon as it is set
Thibault Saunier [Thu, 19 Nov 2020 21:39:33 +0000 (18:39 -0300)]
transcodebin: Add filter as soon as it is set

Instead of waiting so that we can simply use a clocksync element as
filter, otherwise we won't know the pipeline is live as it won't
return NO_PREROLL as one would expect in that case.

Adding it right away shouldn't create any issue, both ways are fine.

3 years agouritranscodebin: Add `setup-source` and `element-setup` signals
Thibault Saunier [Thu, 19 Nov 2020 21:29:15 +0000 (18:29 -0300)]
uritranscodebin: Add `setup-source` and `element-setup` signals

The same way as playbinX does it as it is often quite useful

3 years agotranscode: Port to encodebin2
Thibault Saunier [Thu, 19 Nov 2020 20:55:10 +0000 (17:55 -0300)]
transcode: Port to encodebin2

This allows supporting muxing sinks like hlssink2 or splitmux

3 years agotranscoder: Handle the case where several errors are posted
Thibault Saunier [Thu, 19 Nov 2020 20:55:10 +0000 (17:55 -0300)]
transcoder: Handle the case where several errors are posted

There were cases where the loop was already destroyed when we were
receiving the following message.

3 years agotranscoder: Minor refactoring to output better debug logs
Thibault Saunier [Thu, 19 Nov 2020 20:54:28 +0000 (17:54 -0300)]
transcoder: Minor refactoring to output better debug logs

3 years agohlssink2: Mark as Muxer
Thibault Saunier [Thu, 19 Nov 2020 20:51:56 +0000 (17:51 -0300)]
hlssink2: Mark as Muxer

The way it is usable by encodebin2. This is what splitmux does already.

3 years agova: decoder: Picture dups only holds GstBuffer
Víctor Manuel Jáquez Leal [Mon, 30 Nov 2020 16:12:14 +0000 (17:12 +0100)]
va: decoder: Picture dups only holds GstBuffer

Also removes the warning log message at destroying buffers when picture free()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1841>

3 years agova: Remove gst_va_decoder_destroy_buffers()
Víctor Manuel Jáquez Leal [Mon, 30 Nov 2020 14:01:01 +0000 (15:01 +0100)]
va: Remove gst_va_decoder_destroy_buffers()

Since GstVaDecodePicture is destroyed completely with its free() function and
it's used as destroy notify by codecs picture, there's no need to call
gst_va_decoder_destroy_buffers() externally, since the codecs base classes
destroy the codec picture when it's required.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1841>

3 years agova: Destroy picture unreleased buffers when finalize.
He Junyan [Thu, 26 Nov 2020 06:04:31 +0000 (14:04 +0800)]
va: Destroy picture unreleased buffers when finalize.

The current way of GstVaDecodePicture's finalize will leak some
resource such as parameter buffers and slice data.
The current way deliberately leaves these resource releasing logic
to va decoder related function and trigger a warning if we free the
GstVaDecodePicture without releasing these resources.
But in practice, sometimes, you do not have the chance to release
these resource before picture is freed. For example, H264/Mpeg2
support multi slice NALs/Packets for one frame. It is possible that
we already succeed to parse and generate the first several slices
data by _decode_slice(), but then we get a wrong slice NAL/packet
and fail to parse it. We decide to discard the whole frame in the
decoder's base class, it just free the current picture and does not
trigger sub class's function again. In this kind of cases, we do
not have the chance to cleanup the resource, and the resource will
be leaked.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1841>

3 years agoqroverlay: Reuse the same OverlayComposition object when possible
Thibault Saunier [Sat, 21 Nov 2020 22:00:02 +0000 (19:00 -0300)]
qroverlay: Reuse the same OverlayComposition object when possible

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1829>

3 years agoqroverlay: Rework basing it on overlaycomposition
Thibault Saunier [Fri, 20 Nov 2020 14:28:25 +0000 (11:28 -0300)]
qroverlay: Rework basing it on overlaycomposition

The base class is now a bin which wraps the `overlaycomposition`
element and implements the `draw` signal.

This way we support all the video formats the GstVideoOverlayComposition
API supports and the blending code can be reused. It is also possible
to have the blending happen in the sinks now.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1829>

3 years agod3d11h264dec: Reconfigure decoder object on DPB size change
Seungha Yang [Wed, 25 Nov 2020 20:55:29 +0000 (05:55 +0900)]
d3d11h264dec: Reconfigure decoder object on DPB size change

Even if resolution and/or bitdepth is not updated, required
DPB size can be changed per SPS update and it could be even
larger than previously configured size of DPB. If so, we need
to reconfigure DPB d3d11 texture pool again.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1839>

3 years agoaudio: Use new AudioFormatInfo::fill_silence function
Marijn Suijten [Wed, 25 Nov 2020 16:52:42 +0000 (17:52 +0100)]
audio: Use new AudioFormatInfo::fill_silence function

The function is renamed to be properly associated with AudioFormatInfo
(its instance) instead of AudioFormat (an unrelated enum), see [1] for
the rename itself.

[1]: https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/940

3 years agoplayer: Fix get_current_subtitle_track annotation
Philippe Normand [Thu, 5 Nov 2020 17:14:22 +0000 (17:14 +0000)]
player: Fix get_current_subtitle_track annotation

As the info returned is a new object, the annotation should be transfer-full,
similarly to the get_current_{audio,video}_track() implementations.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1775>

3 years agova: allocator: add a memory pool object helper
Víctor Manuel Jáquez Leal [Mon, 23 Nov 2020 19:44:27 +0000 (20:44 +0100)]
va: allocator: add a memory pool object helper

Since both allocators use a memory pool, with its mutex and cond, this patch
refactors it into a single internal object, implementing a generic GstMemory
pool.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1815>

3 years agova: pool, allocator: honor GST_BUFFER_POOL_ACQUIRE_FLAG_DONTWAIT
Víctor Manuel Jáquez Leal [Tue, 17 Nov 2020 13:53:05 +0000 (14:53 +0100)]
va: pool, allocator: honor GST_BUFFER_POOL_ACQUIRE_FLAG_DONTWAIT

In order to honor GST_BUFFER_POOL_ACQUIRE_FLAG_DONTWAIT in VA pool, allocators'
wait_for_memory() has to be decoupled from their prepare_buffer() so it could be
called in pools' acquire_buffer() if the flag is not set.

wait_for_memory() functions are blocking so the received memories are assigned
to the fist requested buffer, if multithreaded calls. For this a new mutex were
added.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1815>

3 years agova: allocator: broadcast when flushing
Víctor Manuel Jáquez Leal [Tue, 17 Nov 2020 12:18:37 +0000 (13:18 +0100)]
va: allocator: broadcast when flushing

This patch handles when the bufferpool request a new buffer while
flushing.

Also fixes the usage of g_cond_wait(), which demands to be used
inside a loop to avoid spurious wakeups.

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1815>

3 years agova: allocator: free allocator when a mem is held
Víctor Manuel Jáquez Leal [Tue, 17 Nov 2020 12:17:03 +0000 (13:17 +0100)]
va: allocator: free allocator when a mem is held

An application, using for example appsink, can hold buffers from any
va allocator after setting the pipeline to NULL. We need to destroy
the allocator when that memory is unrefed.

This patch juggles a bit with the allocator reference count in
memories in order to achieve this:

1. When memory is created no alloc ref is modified
2. When memory is released, alloc ref is decreased
3. When memory is reassiged to a buffer, alloc ref is increased
4. When memory is flushed, alloc ref is increased becase it is going
   to be decreased in gst_memory_unref()

Also this patch moves the deallocation of member variables to
finalize() rather than dispose()

Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-bad/-/merge_requests/1815>