platform/upstream/gstreamer.git
5 years agoFix zlib detection when there is no pkg-config file
Xavier Claessens [Mon, 19 Nov 2018 21:31:20 +0000 (16:31 -0500)]
Fix zlib detection when there is no pkg-config file

5 years agoaudiodecoder: add API for setting caps on the source pad
Tomasz Andrzejak [Tue, 13 Nov 2018 16:40:23 +0000 (17:40 +0100)]
audiodecoder: add API for setting caps on the source pad

This patch adds API in the audio decoder base class for setting the arbitrary
caps on the source pad.  Previously only caps converted from audio info were
possible.  This is particularly useful when subclass wants to set caps features
for audio decoder producing metadata.

5 years agogl: Fix subclassing of GstGLContextEGL
Xavier Claessens [Wed, 17 Oct 2018 19:58:40 +0000 (15:58 -0400)]
gl: Fix subclassing of GstGLContextEGL

5 years agovideo-format: Move Y210 format declaration to avoid ABI break
Philippe Normand [Thu, 15 Nov 2018 09:28:11 +0000 (11:28 +0200)]
video-format: Move Y210 format declaration to avoid ABI break

The Y210 format was added in the middle of the formats enum and list,
introducing an ABI break.

This issue was detected thanks to the gstreamer-rs test harness.

5 years agortpbasepayload: Update current seqnum for buffer lists
Linus Svensson [Tue, 16 Oct 2018 10:31:57 +0000 (12:31 +0200)]
rtpbasepayload: Update current seqnum for buffer lists

The current sequence number will be the one from the first RTP buffer
when a buffer list is pushed, but should be the last one.

Fixes #495

5 years agomeson: Cleanup old FIXMEs that relied on meson bugfixes
Nirbheek Chauhan [Sun, 28 Oct 2018 13:47:04 +0000 (13:47 +0000)]
meson: Cleanup old FIXMEs that relied on meson bugfixes

5 years agovideo-anc: Add test for VBI encoder
Sebastian Dröge [Wed, 7 Nov 2018 14:45:21 +0000 (16:45 +0200)]
video-anc: Add test for VBI encoder

5 years agovideo-anc: Implement a VBI encoder
Sebastian Dröge [Wed, 7 Nov 2018 13:12:13 +0000 (15:12 +0200)]
video-anc: Implement a VBI encoder

This allows writing out data from caption meta and similar to VBI

5 years agovideo-anc: Add comment about our assumption of the ADF user data format
Sebastian Dröge [Fri, 9 Nov 2018 08:54:57 +0000 (10:54 +0200)]
video-anc: Add comment about our assumption of the ADF user data format

We assume here the same data format for the user data as for the
DID/SDID: 10 bits with parity in the upper 2 bits. In theory some
standards could define this differently and even have full 10 bits of
user data but there does not seem to be a single such standard after
all these years.

5 years agovideo-anc: Add support for parsing composite ADF and check the packets' checksum
Sebastian Dröge [Thu, 8 Nov 2018 16:15:58 +0000 (18:15 +0200)]
video-anc: Add support for parsing composite ADF and check the packets' checksum

And add the checksum to the data we test in the unit test.

5 years agovideo-anc: Fix documentation about SDID_block_number field
Sebastian Dröge [Wed, 7 Nov 2018 12:20:19 +0000 (14:20 +0200)]
video-anc: Fix documentation about SDID_block_number field

It was giving type 2 for both variants.

5 years agovideo-anc: Add invalid-argument guards to public GstVideoVBIParser API
Sebastian Dröge [Wed, 7 Nov 2018 11:58:17 +0000 (13:58 +0200)]
video-anc: Add invalid-argument guards to public GstVideoVBIParser API

5 years agoAdd Gitlab CI configuration
Jordan Petridis [Mon, 12 Nov 2018 11:55:24 +0000 (13:55 +0200)]
Add Gitlab CI configuration

This commit adds a .gitlab-ci.yml file, which uses a feature
to fetch the config from a centralized repository. The intent is
to have all the gstreamer modules use the same configuration.

The configuration is currently hosted at the gst-ci repository
under the gitlab/ci_template.yml path.

Part of https://gitlab.freedesktop.org/gstreamer/gstreamer-project/issues/29

5 years agogl: Make sure that GST_GLSL_PROFILE_ANY gets the correct value in introspection
Sebastian Dröge [Sun, 11 Nov 2018 06:52:27 +0000 (08:52 +0200)]
gl: Make sure that GST_GLSL_PROFILE_ANY gets the correct value in introspection

Currently in Python it would become a signed 64 bit value but should
actually be an unsigned 32 bit value with all bits set.

This is the same problem as with GST_MESSAGE_TYPE_ANY.

See https://bugzilla.gnome.org/show_bug.cgi?id=732633

5 years agoglupload: dmabuf: be explicit about gl formats used
Sjoerd Simons [Fri, 9 Nov 2018 08:32:49 +0000 (09:32 +0100)]
glupload: dmabuf: be explicit about gl formats used

Rather then letting gst_gl_memory_setup_buffer guess the GL format used
for an eglimage after importing a dmabuf be explicit about it. This
fixes issues where dmabuf import may have used another format then
gst_gl_format_from_video_info would guess on the basis of the available
GL extensions.

In particular on etnaviv the gst_gl_format_from_video_info would
assuming a luminance + alpha GL format is used for YUY2, but the dmabuf
import will always use RG88. Which causes images to end up somewhat pink when
displayed on the screen.

5 years agogl/egl: Determine correct format on dmabuf import
Sjoerd Simons [Fri, 9 Nov 2018 09:30:06 +0000 (10:30 +0100)]
gl/egl: Determine correct format on dmabuf import

When importing an egl image from dmabuf gst_gl_format_from_video_info
was used to work what the result GL format will be. Unfortunately that
will only work if the conventional format and the choosen DRM fourcc for
the format match up.

On etnaviv platforms there is no support for GL_EXT_texture_rg, so the
GL format chosen for YUY2 ends up being GST_GL_LUMINANCE_ALPHA. However
DRM does not do luminance + alpha as it's a legacy GL thing, so the
dmabuf import ends up using DRM_FORMAT_GR88.

To fix this, tie the DRM_FORMAT and the GL format together so they
always match up.

5 years agourisourcebin: Avoid potential unitialized/wrong bitrate value
Edward Hervey [Fri, 9 Nov 2018 11:22:31 +0000 (12:22 +0100)]
urisourcebin: Avoid potential unitialized/wrong bitrate value

Only calculate and set the bitrate if all conditions are met.

5 years agourisourcebin: Fix previous commit
Edward Hervey [Wed, 7 Nov 2018 15:28:28 +0000 (16:28 +0100)]
urisourcebin: Fix previous commit

rebase wasn't correct :)

5 years agotests/urisourcebin: add get/set watermark test
Matthew Waters [Mon, 30 Jul 2018 07:56:18 +0000 (17:56 +1000)]
tests/urisourcebin: add get/set watermark test

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/60

5 years agourisourcebin: add a statistics property for queueing
Matthew Waters [Mon, 30 Jul 2018 04:57:23 +0000 (14:57 +1000)]
urisourcebin: add a statistics property for queueing

It contains the minimum/maximum/average byte and time levels of the
queues inside this urisourcebin

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/60

5 years agouisourcebin: update forwarded properties on change
Matthew Waters [Mon, 30 Jul 2018 04:56:40 +0000 (14:56 +1000)]
uisourcebin: update forwarded properties on change

Allows changing queueing properties at runtime

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/60

5 years agourisourcebin: add low/high-watermark properties
Matthew Waters [Mon, 30 Jul 2018 04:48:35 +0000 (14:48 +1000)]
urisourcebin: add low/high-watermark properties

These are forwarded to the contained queues.

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/60

5 years agourisourcebin: limit the byte size of the queue based on the buffer-size
Matthew Waters [Mon, 23 Jul 2018 04:14:32 +0000 (14:14 +1000)]
urisourcebin: limit the byte size of the queue based on the buffer-size

Use the bitrate advertised by queue2 to determine the limits to
set across possibly multiple queue2/downloadbuffer elements.  e.g.
with two queue2's and a max-bytes based on the ratio of the
bitrate/cumulative_bitrate multiplied by the buffer_size set on urisourcebin.

This allows finer grained control over the buffer used by all the queue
elements inside urisourcebin.  Instead of a maximum of
n_streams*buffer_size being used, only buffer_size will be used however
we will fallback to n_streams*buffer_size if one of the queue2's does
not have bitrate information.

https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/60

5 years agourisourcebin: remove unused instance variable streams
Matthew Waters [Wed, 27 Jun 2018 10:17:08 +0000 (20:17 +1000)]
urisourcebin: remove unused instance variable streams

5 years agoexamples: Fix libgstvideo linking
Edward Hervey [Mon, 5 Nov 2018 11:00:55 +0000 (12:00 +0100)]
examples: Fix libgstvideo linking

Use the local libgstvideo (and not the system wide one)

5 years agoaudio: const gpointer is not the same as gconstpointer/const void *
Sebastian Dröge [Sun, 4 Nov 2018 19:17:28 +0000 (21:17 +0200)]
audio: const gpointer is not the same as gconstpointer/const void *

See https://bugzilla.gnome.org/show_bug.cgi?id=664491

5 years agoUpdate common submodule location without /git/ directory
Matthew Waters [Mon, 5 Nov 2018 05:35:12 +0000 (05:35 +0000)]
Update common submodule location without /git/ directory

5 years agoClone the code from gitlab
Haihao Xiang [Mon, 5 Nov 2018 04:06:15 +0000 (12:06 +0800)]
Clone the code from gitlab

This fixes https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/503

5 years agovideo-anc: Add unit tests for VBI parsing
Sebastian Dröge [Thu, 1 Nov 2018 17:19:51 +0000 (19:19 +0200)]
video-anc: Add unit tests for VBI parsing

https://bugzilla.gnome.org/show_bug.cgi?id=797363

5 years agovideo-anc: Fix bounds checks when parsing VBI data
Sebastian Dröge [Thu, 1 Nov 2018 17:19:03 +0000 (19:19 +0200)]
video-anc: Fix bounds checks when parsing VBI data

We were reading more bytes than we allocated.

https://bugzilla.gnome.org/show_bug.cgi?id=797363

5 years agoglupload: Only renegotiate if the caps are incompatible
Nicolas Dufresne [Wed, 31 Oct 2018 18:30:14 +0000 (18:30 +0000)]
glupload: Only renegotiate if the caps are incompatible

There is new code that ensures that we renegotiate after an
uploader transition if the negotiated caps have changed.

The problem is that the raw uploader will not really try and
fixate the input caps, but instead of return a subset with the
only the supported target texture.

This had two effect, raw uploads was always done renegotiated
once and the raw upload unit test was now failing as it didn't
expect a renegotiation.

As it's a valid check, simply relax the gst_caps_is_equal() check
and use a gst_caps_is_subset() instead.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload: Do prepend the preferred caps
Nicolas Dufresne [Wed, 31 Oct 2018 18:26:42 +0000 (18:26 +0000)]
glupload: Do prepend the preferred caps

The direct dmabuf upload does color conversion, so when it transforms
the caps, it replaces the format with all formats found through the
format query. When this uploader can't be used, it makes the upstream
source pick a unsupported format.

To fix this, we only append the caps with a list of format. So the
source will only pick one of these formats if the downstream preferred
format is not supported. A negotiation failure after this would be
normal.

This fixes pipelines without a glcolorconvert element.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload-test: Don't use gboolean to store enums
Nicolas Dufresne [Wed, 31 Oct 2018 18:24:33 +0000 (18:24 +0000)]
glupload-test: Don't use gboolean to store enums

The unit test makes mixed usage of ret value. Sometimes its does
stores an enum and at other moment a boolean. Also fix test
using boolean instead of the correct enum value.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoexamples/playback-test: New entry for text-offset updates
Philippe Normand [Sun, 28 Oct 2018 14:46:15 +0000 (14:46 +0000)]
examples/playback-test: New entry for text-offset updates

https://bugzilla.gnome.org/show_bug.cgi?id=797134

5 years agoplaybin3: New text-offset property
Philippe Normand [Sun, 28 Oct 2018 14:45:26 +0000 (14:45 +0000)]
playbin3: New text-offset property

This new property controls the synchronisation offset between the text and video
streams. Positive values make the text ahead of the video and negative values
make the text go behind the video.

https://bugzilla.gnome.org/show_bug.cgi?id=797134

5 years agoplaybin: New text-offset property
Philippe Normand [Sun, 28 Oct 2018 14:43:01 +0000 (14:43 +0000)]
playbin: New text-offset property

This new property controls the synchronisation offset between the text and video
streams. Positive values make the text ahead of the video and negative values
make the text go behind the video.

https://bugzilla.gnome.org/show_bug.cgi?id=797134

5 years agoplaysink: Add text-offset property
Philippe Normand [Sun, 28 Oct 2018 14:42:31 +0000 (14:42 +0000)]
playsink: Add text-offset property

When the playsink contains a text chain this property controls the
synchronisation of the subtitles and video by controlling the underlying
subtitleoverlay::subtitle-ts-offset property.

https://bugzilla.gnome.org/show_bug.cgi?id=797134

5 years agosubtitleoverlay: Add a subtitle-ts-offset property
Philippe Normand [Sun, 28 Oct 2018 14:35:17 +0000 (14:35 +0000)]
subtitleoverlay: Add a subtitle-ts-offset property

This property controls the synchronisation offset between text and video in
nanoseconds, by updating the parser src pad offset.

https://bugzilla.gnome.org/show_bug.cgi?id=797134

5 years agovideo: add Y210 pixel format.
Wangfei [Tue, 9 Oct 2018 07:25:37 +0000 (15:25 +0800)]
video: add Y210 pixel format.

This pixel format is packed format with 4:2:2 sample and 10
available bits of each channel.

https://bugzilla.gnome.org/show_bug.cgi?id=797267

5 years agoglupload: dmabuf-direct: report driver limitations to debug log
Philipp Zabel [Wed, 17 Oct 2018 15:22:10 +0000 (17:22 +0200)]
glupload: dmabuf-direct: report driver limitations to debug log

Report in the DEBUG log if the driver does not support importing a given
format with linear modifiers non-externally.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload: dmabuf-direct: query formats before modifiers
Philipp Zabel [Wed, 17 Oct 2018 15:11:20 +0000 (17:11 +0200)]
glupload: dmabuf-direct: query formats before modifiers

The EXT_image_dma_buf_import_modifiers extension [1] states regarding
eglQueryDmaBufModifiersEXT:

    The format must be one of those returned by the
    eglQueryDmaBufFormatsEXT command.

To comply with this requirement eglQueryDmaBufFormatsEXT must be called
before eglQueryDmaBufModifiersEXT.

[1] https://www.khronos.org/registry/EGL/extensions/EXT/EGL_EXT_image_dma_buf_import_modifiers.txt

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload: debug output from dmabuf and dmabuf_direct upload transform_caps
Philipp Zabel [Wed, 17 Oct 2018 15:09:26 +0000 (17:09 +0200)]
glupload: debug output from dmabuf and dmabuf_direct upload transform_caps

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload: calculate DRM fourcc once for direct dmabuf upload
Philipp Zabel [Wed, 17 Oct 2018 13:47:07 +0000 (15:47 +0200)]
glupload: calculate DRM fourcc once for direct dmabuf upload

Calculate DRM fourcc and report to the DEBUG log about it only once
instead of three times in gst_egl_image_from_dmabuf_direct().

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload: Implement direct dmabuf uploader
Michael Olbrich [Thu, 5 Jul 2018 08:17:27 +0000 (10:17 +0200)]
glupload: Implement direct dmabuf uploader

The idea is that some GPUs (like the Vivante series) can actually
perform the YUV->RGB conversion internally, so no custom conversion
shaders are needed. To make use of this feature, we need an additional
uploader that can import DMABUF FDs and also directly pass the pixel
format, relying on the GPU to do the conversion.

Based on patches from Nicolas Dufresne <nicolas.dufresne@collabora.com> and
Carlos Rafael Giani <dv@pseudoterminal.org>.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agogluploadelement: try to avoid dropping buffers
Michael Olbrich [Wed, 4 Jul 2018 07:29:57 +0000 (09:29 +0200)]
gluploadelement: try to avoid dropping buffers

Without this, a buffer is dropped if glupload indicates that it is
necessary to reconfigure.
Avoid this by explicitly reconfiguring immediately and uploading the buffer
again.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload: handle upload methods with different caps
Michael Olbrich [Wed, 4 Jul 2018 07:21:00 +0000 (09:21 +0200)]
glupload: handle upload methods with different caps

If a upload method is selected then use it exclusively in transform_caps().
Also, reconfigure if the current caps don't match the current upload
method.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload: allow system memory for dmabuf in transform_caps
Michael Olbrich [Thu, 5 Jul 2018 13:17:30 +0000 (15:17 +0200)]
glupload: allow system memory for dmabuf in transform_caps

This should not be necessary, but currently not all plugins that provide
dmabuf memory announce this with caps features, e.g. v4l2.
The static caps already contain the system memory. It didn't break before
because other upload methods provide the necessary transformation.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglupload: try to use the last method after reconfigure
Michael Olbrich [Wed, 4 Jul 2018 07:17:11 +0000 (09:17 +0200)]
glupload: try to use the last method after reconfigure

Reconfigure will trigger a set_caps which clears the upload method.
Remember the method in this case and start with it.
Wrap around once to try all methods if necessary.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agogl/egl: Add gst_egl_image_from_dmabuf_direct() function
Michael Olbrich [Tue, 30 Oct 2018 11:25:02 +0000 (12:25 +0100)]
gl/egl: Add gst_egl_image_from_dmabuf_direct() function

The colorspace conversion happens during the upload so the necessary hints
must be provided to ensure that the conversion works correctly.

At least the Mesa Intel driver will create a texture without error but
produces an incorrect result. Use eglQueryDmaBufModifiersEXT() to check if
non-external upload is supported for the given format.

Based on a patch from Carlos Rafael Giani <dv@pseudoterminal.org>.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agoglmemory: Fix n_wrapped_pointers usage
Nicolas Dufresne [Wed, 7 Jun 2017 16:00:08 +0000 (18:00 +0200)]
glmemory: Fix n_wrapped_pointers usage

gst_gl_memory_setup_buffer() was not properly using the number
of pointers to wrapped. This also fixes the validation, as we
only support 1 wrapper per view, or num_planes * views wrapper.

https://bugzilla.gnome.org/show_bug.cgi?id=783521

5 years agooverlaycomposition: Fix implicit use of memset
Johan Bjäreholt [Mon, 29 Oct 2018 08:18:35 +0000 (09:18 +0100)]
overlaycomposition: Fix implicit use of memset

https://bugzilla.gnome.org/show_bug.cgi?id=797353

5 years agooverlaycomposition: Add plugin inspect file
Sebastian Dröge [Sun, 28 Oct 2018 17:41:38 +0000 (17:41 +0000)]
overlaycomposition: Add plugin inspect file

5 years agoexamples: overlaycomposition: Fix Meson build
Philippe Normand [Sun, 28 Oct 2018 17:21:08 +0000 (17:21 +0000)]
examples: overlaycomposition: Fix Meson build

5 years agooverlaycomposition: New element that allows applications to draw GstVideoOverlayCompo...
Sebastian Dröge [Mon, 1 Oct 2018 15:10:34 +0000 (18:10 +0300)]
overlaycomposition: New element that allows applications to draw GstVideoOverlayComposition on a stream

https://bugzilla.gnome.org/show_bug.cgi?id=797234

5 years agoparsebin: Add missing locks/unlocks of the chain mutex
Sebastian Dröge [Wed, 8 Aug 2018 15:35:08 +0000 (17:35 +0200)]
parsebin: Add missing locks/unlocks of the chain mutex

Before freeing pending pads it is required to hold the mutex, that's
what is protecting the list of pending pads in other places.

https://bugzilla.gnome.org/show_bug.cgi?id=796935

5 years agotests: Solidify tcp connection check
Edward Hervey [Sat, 27 Oct 2018 13:48:13 +0000 (15:48 +0200)]
tests: Solidify tcp connection check

The previous failure was a timeout which was due to the sending pipeline
pushing test buffer *before* the remote client was accepted. We would
therefore never get the buffer on the other side.

While the client socket would indeed appear as "connected", this doesn't
mean that the remote server side did "accept" it (which is where we then
add it to the list of remote parties to which data will be sent).

The problem isn't with the element implementation, but to the nature of
TCP 3-way handshake.

In order to make the test reliable, wait for the sink to have accepted
the remote client (by checking the number of handles) before sending out
test buffers.

5 years agodecodebin3: fix eos event leak
Roland Jon [Tue, 24 Jul 2018 10:40:36 +0000 (18:40 +0800)]
decodebin3: fix eos event leak

https://bugzilla.gnome.org/show_bug.cgi?id=796860

5 years agooggdemux: Answer POSITION query
Sebastian Dröge [Wed, 17 Oct 2018 16:43:51 +0000 (19:43 +0300)]
oggdemux: Answer POSITION query

5 years agomeson: Replace empty configuration_data() with copy keyword
Seungha Yang [Thu, 27 Sep 2018 12:21:14 +0000 (21:21 +0900)]
meson: Replace empty configuration_data() with copy keyword

Use 'copy' keyword to avoid meson warning message.
Note that 'copy' keyword in configure_file() is available
since meson 0.47.0

https://bugzilla.gnome.org/show_bug.cgi?id=797298

5 years agovideometa: Log map errors with GST_ERROR, not DEBUG
Nirbheek Chauhan [Mon, 15 Oct 2018 14:15:46 +0000 (19:45 +0530)]
videometa: Log map errors with GST_ERROR, not DEBUG

Otherwise it's very easy to miss them when gst_video_frame_map() fails

5 years agortspconnection: do not duplicate authentication headers
Mathieu Duponchelle [Wed, 10 Oct 2018 21:15:49 +0000 (23:15 +0200)]
rtspconnection: do not duplicate authentication headers

rtsp_connection_send takes care of adding those already,
and some reverse proxies such as nginx will reject the request
altogether if the Authorization header is present twice,
even with the same value.

https://bugzilla.gnome.org/show_bug.cgi?id=797272

5 years agortpbasepayload: rtpbasedepayload: Add source-info property
Stian Selnes [Wed, 10 Oct 2018 18:16:12 +0000 (14:16 -0400)]
rtpbasepayload: rtpbasedepayload: Add source-info property

Add a source-info property that will read/write meta to the buffers
about RTP source information. The GstRTPSourceMeta can be used to
transport information about the origin of a buffer, e.g. the sources
that is included in a mixed audio buffer.

A new function gst_rtp_base_payload_allocate_output_buffer() is added
for payloaders to use to allocate the output RTP buffer with the correct
number of CSRCs according to the meta and fill it.

RTPSourceMeta does not make sense on RTP buffers since the information
is in the RTP header. So the payloader will strip the meta from the
output buffer.

https://bugzilla.gnome.org/show_bug.cgi?id=761947

5 years agomeson: use new 'python' module instead of deprecated
Tim-Philipp Müller [Mon, 8 Oct 2018 00:07:45 +0000 (01:07 +0100)]
meson: use new 'python' module instead of deprecated

https://github.com/mesonbuild/meson/pull/4169

5 years agoexamples: gl: remove bitrotten clutter examples
Tim-Philipp Müller [Sun, 7 Oct 2018 18:09:24 +0000 (19:09 +0100)]
examples: gl: remove bitrotten clutter examples

https://bugzilla.gnome.org/show_bug.cgi?id=797173

5 years agotests: audioconvert: test caps fixate with layout conversion
George Kiagiadakis [Fri, 5 Oct 2018 12:37:45 +0000 (15:37 +0300)]
tests: audioconvert: test caps fixate with layout conversion

https://bugzilla.gnome.org/show_bug.cgi?id=797225

5 years agoaudioconvert: remove layout from structure when fixating caps
Matej Knopp [Sat, 29 Sep 2018 08:39:46 +0000 (10:39 +0200)]
audioconvert: remove layout from structure when fixating caps

otherwise caps intersection always fails when converting non-interleaved to interleaved audio

https://bugzilla.gnome.org/show_bug.cgi?id=797225

5 years agoRevert "Revert "gl: use correct blend functions for unpremultiplied alpha""
Matthew Waters [Thu, 4 Oct 2018 15:19:29 +0000 (01:19 +1000)]
Revert "Revert "gl: use correct blend functions for unpremultiplied alpha""

This reverts commit 452d90a88c6e68a3daa38e02213782836788236a.

This was actually correct. sorry for the noise!

5 years agoglimagesink: make multiview none equivalent to mono
Matthew Waters [Thu, 4 Oct 2018 14:48:58 +0000 (00:48 +1000)]
glimagesink: make multiview none equivalent to mono

Fixes the internal viewconvert to not scale buffers for output with the
following pipeline:

gltestsrc ! glimagesink

It also fixes overlay composition with a resized output with an OpenGL
upstream:

gltestsrc ! timeoverlay ! glimagesink

5 years agoRevert "gl: use correct blend functions for unpremultiplied alpha"
Matthew Waters [Thu, 4 Oct 2018 14:03:09 +0000 (00:03 +1000)]
Revert "gl: use correct blend functions for unpremultiplied alpha"

This reverts commit 779d028f27b7bc6eb7abe841fe2cda97de3be983.

5 years agogl: use correct blend functions for unpremultiplied alpha
Matthew Waters [Thu, 4 Oct 2018 13:43:19 +0000 (23:43 +1000)]
gl: use correct blend functions for unpremultiplied alpha

Fallback to the closest but incorrect values we had before if we can't
do this correctly.

5 years agogl/overlaycompositor: don't force unpremultiplied alpha
Matthew Waters [Thu, 4 Oct 2018 13:23:03 +0000 (23:23 +1000)]
gl/overlaycompositor: don't force unpremultiplied alpha

Remove testing code from 23159365618409011745b554b4ad371700961121

5 years agogl/overlaycompositor: support both (un)premultiplied alpha
Matthew Waters [Thu, 4 Oct 2018 13:02:05 +0000 (23:02 +1000)]
gl/overlaycompositor: support both (un)premultiplied alpha

Using the correct blend modes for each case or converting to
premultipled in the very unlikely case that separate blend modes are
unavailable on ancient opengl hardware.

5 years agogl: GstGLAPI should be gst_gl_api_ in lower-case
Sebastian Dröge [Thu, 4 Oct 2018 11:50:43 +0000 (14:50 +0300)]
gl: GstGLAPI should be gst_gl_api_ in lower-case

5 years agoglviewconvert: Add backwards compatibility gst_gl_stereo_downmix_mode_get_type()
Sebastian Dröge [Thu, 4 Oct 2018 10:38:59 +0000 (13:38 +0300)]
glviewconvert: Add backwards compatibility gst_gl_stereo_downmix_mode_get_type()

It's proper new name is gst_gl_stereo_downmix_get_type().

5 years agogl: Add gl-enumtypes.[ch] to CLEANFILES
Sebastian Dröge [Thu, 4 Oct 2018 10:33:30 +0000 (13:33 +0300)]
gl: Add gl-enumtypes.[ch] to CLEANFILES

5 years agovideo-overlay-composition: Correctly mark GstVideoOverlayFormatFlags as flags
Sebastian Dröge [Thu, 4 Oct 2018 08:36:09 +0000 (11:36 +0300)]
video-overlay-composition: Correctly mark GstVideoOverlayFormatFlags as flags

And register a GType for it.

5 years agogl/meson: fix wrong overlay filename
Matthew Waters [Thu, 4 Oct 2018 06:45:34 +0000 (16:45 +1000)]
gl/meson: fix wrong overlay filename

5 years agoglimagesink: only update the output info iff there's a valid display_rect
Matthew Waters [Fri, 28 Sep 2018 02:09:16 +0000 (12:09 +1000)]
glimagesink: only update the output info iff there's a valid display_rect

Attempting to use the MAX(1, display_rect) would result in the overlay
composition attempting to draw into 1x1 buffer and calculate some
grossly incorrect sizes.

previously failing case:
gltestsrc ! textoverlay text=GStreamer ! glimagesinkelement

5 years agogl: add a new overlay compositor element
Matthew Waters [Thu, 27 Sep 2018 03:35:15 +0000 (13:35 +1000)]
gl: add a new overlay compositor element

Flattens all the overlays from the GstVideoOverlayCompositionMeta into
the video stream.

https://bugzilla.gnome.org/show_bug.cgi?id=759867

5 years agogl/overlaycompositor: add support for flipping the output
Matthew Waters [Thu, 27 Sep 2018 03:31:09 +0000 (13:31 +1000)]
gl/overlaycompositor: add support for flipping the output

This is required to use gloverlaycompositor outside of glimagesink where
the output is not automatically flipped

5 years agogl/overlaycompositor: handle ANY caps features gracefully
Matthew Waters [Thu, 27 Sep 2018 03:30:35 +0000 (13:30 +1000)]
gl/overlaycompositor: handle ANY caps features gracefully

5 years agoglfilter: don't complete overwrite caps features from the subclass
Matthew Waters [Thu, 27 Sep 2018 03:29:03 +0000 (13:29 +1000)]
glfilter: don't complete overwrite caps features from the subclass

We can get away with ensuring that the memory:GLMemory caps feature is
present in the output caps

5 years agogl: Add gl-enumtypes.h to nodist_HEADERS
Sebastian Dröge [Wed, 3 Oct 2018 20:27:32 +0000 (23:27 +0300)]
gl: Add gl-enumtypes.h to nodist_HEADERS

5 years agoglviewconvert: Create an anonymous enum for GstGLStereoDownmix
Sebastian Dröge [Wed, 3 Oct 2018 20:17:56 +0000 (23:17 +0300)]
glviewconvert: Create an anonymous enum for GstGLStereoDownmix

Older glib-mkenums completely fails to parse it otherwise.

5 years agogl: Rename GST_TYPE_GL_STEREO_DOWNMIX GType macro everywhere
Sebastian Dröge [Wed, 3 Oct 2018 11:49:32 +0000 (14:49 +0300)]
gl: Rename GST_TYPE_GL_STEREO_DOWNMIX GType macro everywhere

The old one still exists but behind GST_DISABLE_DEPRECATED

5 years agogl: Generate enum GTypes via glib-mkenums
Sebastian Dröge [Wed, 3 Oct 2018 10:06:42 +0000 (13:06 +0300)]
gl: Generate enum GTypes via glib-mkenums

https://bugzilla.gnome.org/show_bug.cgi?id=797240
https://bugzilla.gnome.org/show_bug.cgi?id=797144

5 years agoglsl: Fix error quark string to be more consistent
Sebastian Dröge [Tue, 2 Oct 2018 18:01:31 +0000 (21:01 +0300)]
glsl: Fix error quark string to be more consistent

5 years agoglbasememory: Fix error quark string and move documentation to the right place
Sebastian Dröge [Tue, 2 Oct 2018 17:59:01 +0000 (20:59 +0300)]
glbasememory: Fix error quark string and move documentation to the right place

5 years agoencoding-profile: set_restriction should accept null as valid
Thiago Santos [Tue, 2 Oct 2018 00:51:26 +0000 (17:51 -0700)]
encoding-profile: set_restriction should accept null as valid

It was checking for GST_IS_CAPS only and that would fail if the new
restriction caps was NULL and its documentation says it accepts NULL as
valid input.

5 years agobasetextoverlay: Append our rectangle to the upstream composition, not the other...
Sebastian Dröge [Mon, 1 Oct 2018 16:13:39 +0000 (19:13 +0300)]
basetextoverlay: Append our rectangle to the upstream composition, not the other way around

Upstream is supposed to be painted first, and only then our rectangle.

5 years agobasetextoverlay: Minor cleanup of negotiation functions
Sebastian Dröge [Mon, 1 Oct 2018 15:27:22 +0000 (18:27 +0300)]
basetextoverlay: Minor cleanup of negotiation functions

5 years agovideo-overlay-composition: Handle miniobject writability correctly
Sebastian Dröge [Mon, 1 Oct 2018 09:17:42 +0000 (12:17 +0300)]
video-overlay-composition: Handle miniobject writability correctly

By tracking parents and using the proper GstMiniObject API for detecting
writability instead of just the reference count.

https://bugzilla.gnome.org/show_bug.cgi?id=797230

5 years agovideo: Fix annotations for gst_buffer_add_video_meta_full()
Sebastian Dröge [Sat, 29 Sep 2018 08:26:27 +0000 (11:26 +0300)]
video: Fix annotations for gst_buffer_add_video_meta_full()

5 years agotagdemux: Use upstream GST_EVENT_STREAM_START if present
Edward Hervey [Thu, 27 Sep 2018 13:03:10 +0000 (15:03 +0200)]
tagdemux: Use upstream GST_EVENT_STREAM_START if present

If we have an upstream GST_EVENT_STREAM_START, use that one instead
of creating a new one which could be completely different from the
upstream one and drop information (like the stream flags and stream
object).

Only create a new event if we don't already have one from upstream

https://bugzilla.gnome.org/show_bug.cgi?id=797215

5 years agoglviewconvert: wait and set the gl sync meta on buffers
Matthew Waters [Wed, 26 Sep 2018 03:22:14 +0000 (13:22 +1000)]
glviewconvert: wait and set the gl sync meta on buffers

This may be a cause of out-of-place frames when transforming multiview
buffers.

5 years agoglviewconvert: Copy composition meta from the primary buffer to both outputs
Sebastian Dröge [Tue, 25 Sep 2018 13:56:24 +0000 (16:56 +0300)]
glviewconvert: Copy composition meta from the primary buffer to both outputs

Without this, glviewconvert (and thus glimagesink) will drop all overlay
composition metas.

5 years agoglcolorconvert: Don't copy overlay composition meta over to NULL outbufs
Sebastian Dröge [Tue, 25 Sep 2018 13:56:07 +0000 (16:56 +0300)]
glcolorconvert: Don't copy overlay composition meta over to NULL outbufs

5 years agogstaudioutilsprivate: Fix warnings while setting thread priority
Nirbheek Chauhan [Mon, 24 Sep 2018 13:55:28 +0000 (19:25 +0530)]
gstaudioutilsprivate: Fix warnings while setting thread priority

Also use G_OS_WIN32 instead of _WIN32 for clarity.

5 years agolibs: fix API export/import and 'inconsistent linkage' on MSVC
Tim-Philipp Müller [Sat, 28 Apr 2018 13:50:11 +0000 (14:50 +0100)]
libs: fix API export/import and 'inconsistent linkage' on MSVC

For each lib we build export its own API in headers when we're
building it, otherwise import the API from the headers.

This fixes linker warnings on Windows when building with MSVC.

The problem was that we had defined all GST_*_API decorators
unconditionally to GST_EXPORT. This was intentional and only
supposed to be temporary, but caused linker warnings because
we tell the linker that we want to export all symbols even
those from externall DLLs, and when the linker notices that
they were in external DLLS and not present locally it warns.

What we need to do when building each library is: export
the library's own symbols and import all other symbols. To
this end we define e.g. BUILDING_GST_FOO and then we define
the GST_FOO_API decorator either to export or to import
symbols depending on whether BUILDING_GST_FOO is set or not.
That way external users of each library API automatically
get the import.

While we're at it, add new GST_API_EXPORT in config.h and use
that for GST_*_API decorators instead of GST_EXPORT.

The right export define depends on the toolchain and whether
we're using -fvisibility=hidden or not, so it's better to set it
to the right thing directly than hard-coding a compiler whitelist
in the public header.

We put the export define into config.h instead of passing it via the
command line to the compiler because it might contain spaces and brackets
and in the autotools scenario we'd have to pass that through multiple
layers of plumbing and Makefile/shell escaping and we're just not going
to be *that* lucky.

The export define is only used if we're compiling our lib, not by external
users of the lib headers, so it's not a problem to put it into config.h

Also, this means all .c files of libs need to include config.h
to get the export marker defined, so fix up a few that didn't
include config.h.

This commit depends on a common submodule commit that makes gst-glib-gen.mak
add an #include "config.h" to generated enum/marshal .c files for the
autotools build.

https://bugzilla.gnome.org/show_bug.cgi?id=797185

5 years agomeson: Specify encoding to UTF-8 when building with MSVC
Seungha Yang [Fri, 21 Sep 2018 13:31:33 +0000 (22:31 +0900)]
meson: Specify encoding to UTF-8 when building with MSVC

Fix build on some non-US locale Windows systems

Error:
  gst-plugins-base/tools/gst-device-monitor.c(278): error C2001

https://bugzilla.gnome.org/show_bug.cgi?id=797186