platform/upstream/gstreamer.git
8 years agotests: compositor: drop special case for valgrind timeout
Thiago Santos [Tue, 23 Feb 2016 15:42:19 +0000 (12:42 -0300)]
tests: compositor: drop special case for valgrind timeout

The default one is 6 minutes, the test was using 5 minutes so just
resort to using the default.

For the non-valgrind test also use the default 20 secs instead of
reducing it to 6s. No real reason to set a custom value here.

8 years agotests: compositor: add tests for caps queries
Thiago Santos [Tue, 23 Feb 2016 15:17:59 +0000 (12:17 -0300)]
tests: compositor: add tests for caps queries

Verifies that proper caps are returned based on what downstream
restricts.

8 years agovideoaggregator: fix caps queries to allow proper renegotiation
Thiago Santos [Thu, 18 Feb 2016 13:57:51 +0000 (10:57 -0300)]
videoaggregator: fix caps queries to allow proper renegotiation

When caps are already negotiated it should be possible to
select formats other than the one that was negotiated. If downstream
allows alpha video caps and it has already negotiated to a non-alpha
format, caps queries should still return the alpha caps as a possible
format as caps renegotiation can happen.

Includes tests (for compositor) to check that caps queries done after
a caps has been negotiated returns complete results

https://bugzilla.gnome.org/show_bug.cgi?id=757610

8 years agoglmixer: iterator didn't advance in continue statement
Wang Xin-yu (王昕宇) [Wed, 24 Feb 2016 02:45:17 +0000 (10:45 +0800)]
glmixer: iterator didn't advance in continue statement

Leading to a deadlock.

https://bugzilla.gnome.org/show_bug.cgi?id=760873

8 years agogl: error out if the configured GL API is unsupported by our element
Matthew Waters [Mon, 22 Feb 2016 09:49:52 +0000 (20:49 +1100)]
gl: error out if the configured GL API is unsupported by our element

https://bugzilla.gnome.org/show_bug.cgi?id=759801

8 years agoglvideomixer: don't leak pad's vertex buffer on release_pad
Wang Xin-yu (王昕宇) [Thu, 21 Jan 2016 02:40:36 +0000 (10:40 +0800)]
glvideomixer: don't leak pad's vertex buffer on release_pad

https://bugzilla.gnome.org/show_bug.cgi?id=760873

8 years agoglmixer: Remove usage of GstGLMixerFrameData
Matthew Waters [Tue, 16 Feb 2016 14:08:18 +0000 (01:08 +1100)]
glmixer: Remove usage of GstGLMixerFrameData

Subclasses can just iterate over the list of pads themselves

https://bugzilla.gnome.org/show_bug.cgi?id=760873

8 years agoglmixer: don't hold the object lock while calling into GL
Matthew Waters [Wed, 13 Jan 2016 03:41:22 +0000 (14:41 +1100)]
glmixer: don't hold the object lock while calling into GL

Doing so can deadlock between the GL thread and the object lock e.g.
when performing reconfigure events in glimagesink on a resize event.

https://bugzilla.gnome.org/show_bug.cgi?id=760559

8 years agoglvideomixer: fix checker vbo leak
Wang Xin-yu (王昕宇) [Thu, 21 Jan 2016 08:10:48 +0000 (16:10 +0800)]
glvideomixer: fix checker vbo leak

https://bugzilla.gnome.org/show_bug.cgi?id=760925

8 years agoglvideomixer: par may not exist in the caps
Matthew Waters [Fri, 29 Jan 2016 03:03:26 +0000 (14:03 +1100)]
glvideomixer: par may not exist in the caps

Fixes a critical in the gst-validate tests:

gst_structure_fixate_field_nearest_fraction: assertion 'gst_structure_has_field
(structure, field_name)

8 years agovideoaggregator: don't do caps processing that is not overridable
Matthew Waters [Wed, 14 Oct 2015 10:13:57 +0000 (21:13 +1100)]
videoaggregator: don't do caps processing that is not overridable

Allows the subclass to completely override the chosen src caps.

This is needed as videoaggregator generally has no idea exactly
what operation is being performed.

- Adds a fixate_caps vfunc for fixation
- Merges gst_video_aggregator_update_converters() into
  gst_videoaggregator_update_src_caps() as we need some of its info
  for proper caps handling.
- Pass the downstream caps to the update_caps vfunc

https://bugzilla.gnome.org/show_bug.cgi?id=756207

8 years agogl: move control binding proxy implementation from glvideomixer
Matthew Waters [Wed, 27 Jan 2016 08:56:36 +0000 (19:56 +1100)]
gl: move control binding proxy implementation from glvideomixer

Other elements may need to use it's functionality

8 years agoglvideomixer: implement glBlendFunc and glBlendEquation
Matthew Waters [Tue, 12 Jan 2016 07:21:50 +0000 (18:21 +1100)]
glvideomixer: implement glBlendFunc and glBlendEquation

Allows more blending options than just A over B

e.g. frame comparisons are now possible.
  glvideomixer name=m
    sink_0::zorder=0
    sink_1::zorder=1
    sink_1::blend-equation-rgb={subtract,reverse-subtract}
    sink_1::blend-function-src-rgb=src-color
    sink_1::blend-function-dst-rgb=dst-color
  ! glimagesinkelement
  videotestsrc pattern=checkers-4 ! m.sink_0
  videotestsrc pattern=checkers-8 ! m.sink_1

8 years agogldownload: remove helper api from the library
Matthew Waters [Fri, 11 Dec 2015 04:39:57 +0000 (15:39 +1100)]
gldownload: remove helper api from the library

It was never used by anyone and is not needed anymore with the element
and GstGLMemory's transparent support for downloading textures.

8 years agoglviewconvert: remove set_format
Matthew Waters [Tue, 10 Nov 2015 02:52:30 +0000 (13:52 +1100)]
glviewconvert: remove set_format

We need the caps to be able to
1. check the caps features
2. get the requested texture-target on input/output

8 years agoglmixerbin: fix minor leak
Tim-Philipp Müller [Fri, 6 Nov 2015 12:30:12 +0000 (12:30 +0000)]
glmixerbin: fix minor leak

Don't leak removed list node.

8 years agovideoaggregator: use GST_STIME_ARGS for GstClockTimeDiff
Luis de Bethencourt [Mon, 2 Nov 2015 16:58:57 +0000 (16:58 +0000)]
videoaggregator: use GST_STIME_ARGS for GstClockTimeDiff

No need to manually handle negative values of diff, GST_STIME_ARGS does
exactly this.

8 years agoglmemory: add support for rectangle textures
Matthew Waters [Wed, 28 Oct 2015 13:44:26 +0000 (00:44 +1100)]
glmemory: add support for rectangle textures

Add the various tokens/strings for the differnet texture types (2D, rect, oes)

Changes the GLmemory api to include the GstGLTextureTarget in all relevant
functions.

Update the relevant caps/templates for 2D only textures.

8 years agoglvideomixer: Proxy the ignore-eos videoaggregator property as well
Nirbheek Chauhan [Tue, 6 Oct 2015 15:53:11 +0000 (21:23 +0530)]
glvideomixer: Proxy the ignore-eos videoaggregator property as well

Identical to how the z-order property is proxied

8 years agovideoaggregator: fix compilation with older glib version
Tim-Philipp Müller [Tue, 29 Sep 2015 12:31:18 +0000 (13:31 +0100)]
videoaggregator: fix compilation with older glib version

Remove weird use of private gtype defines and fix compilation
with older glib versions such as 2.36.

https://bugzilla.gnome.org/show_bug.cgi?id=755754

8 years agogl: chain up to the parent class for GstElement::set_context
Matthew Waters [Mon, 28 Sep 2015 12:20:29 +0000 (22:20 +1000)]
gl: chain up to the parent class for GstElement::set_context

https://bugzilla.gnome.org/show_bug.cgi?id=705579

8 years agocheck: Add test for videoaggregator sinkpads being sorted by zorder
Nirbheek Chauhan [Sat, 29 Aug 2015 13:46:38 +0000 (19:16 +0530)]
check: Add test for videoaggregator sinkpads being sorted by zorder

https://bugzilla.gnome.org/show_bug.cgi?id=754285

8 years agoDon't throw compiler warnings with G_DISABLE_ASSERT
Jan Schmidt [Thu, 17 Sep 2015 14:20:13 +0000 (00:20 +1000)]
Don't throw compiler warnings with G_DISABLE_ASSERT

Disable code that warns about unused variables when G_DISABLE_ASSERT
is defined, as it is in tarballs and pre-releases.

8 years agogl: Fix GError leaks during failures
Vineeth T M [Thu, 17 Sep 2015 04:35:02 +0000 (13:35 +0900)]
gl: Fix GError leaks during failures

https://bugzilla.gnome.org/show_bug.cgi?id=755140

8 years agovideoaggregator: Fix mixup of running times and segment positions
Sebastian Dröge [Fri, 11 Sep 2015 10:22:51 +0000 (12:22 +0200)]
videoaggregator: Fix mixup of running times and segment positions

We have to queue buffers based on their running time, not based on
the segment position.

Also return running time from GstAggregator::get_next_time() instead of
a segment position, as required by the API.

Also only update the segment position after we pushed a buffer, otherwise
we're going to push down a segment event with the next position already.

https://bugzilla.gnome.org/show_bug.cgi?id=753196

8 years agovideoaggregator: fix caps query to properly handle alpha formats
Thiago Santos [Wed, 9 Sep 2015 22:51:18 +0000 (19:51 -0300)]
videoaggregator: fix caps query to properly handle alpha formats

Only accept alpha if downstream has alpha as well. It could
theoretically accept alpha unconditionally if blending is
properly implemented for handle it but at the moment this
is a missing feature.

Improves the caps query by also comparing with the template
caps to filter by what the subclass supports.

https://bugzilla.gnome.org/show_bug.cgi?id=754465

8 years agocompositor: Ensure all arguments to CLAMP are signed int
Nirbheek Chauhan [Fri, 4 Sep 2015 20:20:41 +0000 (01:50 +0530)]
compositor: Ensure all arguments to CLAMP are signed int

If any of the arguments to CLAMP are unsigned integers, the comparison causes
an automatic conversion of the signed int to unsigned, which causes -1 to become
UINT_MAX and get clamped to the high value of the CLAMP instead of 0.

See 716 at http://c0x.coding-guidelines.com/6.3.1.8.html

Also add a test for this.

https://bugzilla.gnome.org/show_bug.cgi?id=754576

8 years agovideoaggregator: lift restriction of changing pixel-aspect-ratio
Thiago Santos [Wed, 2 Sep 2015 22:16:56 +0000 (19:16 -0300)]
videoaggregator: lift restriction of changing pixel-aspect-ratio

The videoaggregator can convert PAR, there is no reason for
restricting it.

https://bugzilla.gnome.org/show_bug.cgi?id=754291

8 years agocompositor: variables in clamp_rectangle() should be signed
Nirbheek Chauhan [Sat, 29 Aug 2015 11:21:08 +0000 (16:51 +0530)]
compositor: variables in clamp_rectangle() should be signed

x/y/w/h are signed integers. As can be seen in GstCompositorPad.
The prototype for clamp_rectangle was wrong. This commit reverts the change
and fixes the prototype.

This reverts commit bca444ea4a84c39e9989681f892f6e4cb2033cf9.

8 years agocompositor: remove check for below zero for unsigned value
Luis de Bethencourt [Fri, 28 Aug 2015 14:21:11 +0000 (15:21 +0100)]
compositor: remove check for below zero for unsigned value

CLAMP checks both if value is '< 0' and '> max'. Value will never be a negative
number since it is an unsigned integer. Removing that check and only checking if
it is bigger than max by using MIN().

CID 1320707

8 years agocompositor: Actually use the output resolution for clamping
Nirbheek Chauhan [Wed, 26 Aug 2015 10:10:16 +0000 (15:40 +0530)]
compositor: Actually use the output resolution for clamping

The obscured check in compositor was using the dimensions of the pad to clamp
the h/w of the pad instead of the output resolution, and was doing an incorrect
calculation to do so. Fix that by simplifying the whole calculation by using
corner coordinates. Also add a test for this bug which fell through the cracks,
and just skip all the obscured tests if the pad's alpha is 0.0.

https://bugzilla.gnome.org/show_bug.cgi?id=754107

8 years agovideoaggregator: Always set the pad's buffer_vinfo when storing a buffer
Ben Browitt [Thu, 20 Aug 2015 11:11:56 +0000 (14:11 +0300)]
videoaggregator: Always set the pad's buffer_vinfo when storing a buffer

Otherwise it might be unset, and then the buffer is used and
gst_video_frame_map() will crash because of invalid video-info.

https://bugzilla.gnome.org/show_bug.cgi?id=753805

8 years agoglvideomixer: swap control binding proxy
Matthew Waters [Mon, 3 Aug 2015 09:48:16 +0000 (19:48 +1000)]
glvideomixer: swap control binding proxy

The ref_object and object parameters were the wrong way around.

For the typical use case where an application is setting a
GstControlBinding on the returned ghost pad:
1. our control binding would be removed when the new one was set
2. sync_values calls were not being forwarded from the internal
   pad to the ghost pad.

If an application attempts to perform other control binding
operations (get_* family of functions) on the internal pad, they
will also be forwarded to the ghost pad where a possible
GstControlBinding will provide the necessary values.

8 years agovideoaggregator: Remove broken _clip vfunc
Olivier Crête [Sat, 7 Mar 2015 02:32:04 +0000 (21:32 -0500)]
videoaggregator: Remove broken _clip vfunc

It never does anything.

https://bugzilla.gnome.org/show_bug.cgi?id=745768

8 years agoglstereomix: remove redundant initialization
Luis de Bethencourt [Wed, 29 Jul 2015 16:48:58 +0000 (17:48 +0100)]
glstereomix: remove redundant initialization

v is initialized in the for loop init, no need to do it twice. Removing
first initialization.

8 years agocompositor: Add unit tests for the new aggregator start-time-selection property
Sebastian Dröge [Tue, 28 Jul 2015 19:30:09 +0000 (22:30 +0300)]
compositor: Add unit tests for the new aggregator start-time-selection property

https://bugzilla.gnome.org/show_bug.cgi?id=749966

8 years agoaggregator: Add property to select how to decide on a start time
Sebastian Dröge [Mon, 15 Jun 2015 16:30:20 +0000 (18:30 +0200)]
aggregator: Add property to select how to decide on a start time

Before aggregator based elements always started at running time 0,
now it's possible to select the first input buffer running time or
explicitly set a start-time value.

https://bugzilla.gnome.org/show_bug.cgi?id=749966

8 years agoglvideomixer: Add GstControlBinding proxy
Olivier Crête [Fri, 3 Jul 2015 00:10:50 +0000 (20:10 -0400)]
glvideomixer: Add GstControlBinding proxy

This is used to proxy GstControlBinding to the pad on the
parent object. This avoid having to sync the values in the proxy pad,
this is too early if you have a queue between the pad and the actual
aggregation operation.

https://bugzilla.gnome.org/show_bug.cgi?id=734060

9 years agogl: use gst_gl_display_create_context in more elements.
Julien Isorce [Fri, 19 Jun 2015 10:57:06 +0000 (11:57 +0100)]
gl: use gst_gl_display_create_context in more elements.

glbasefilter, glbasemixer and gltestsrc.

https://bugzilla.gnome.org/show_bug.cgi?id=750310

9 years agovideoaggregator: Remove pixel-aspect-ratio field from the caps returned by getcaps
Thibault Saunier [Mon, 6 Jul 2015 16:51:07 +0000 (18:51 +0200)]
videoaggregator: Remove pixel-aspect-ratio field from the caps returned by getcaps

Avoiding not negotiated error while negotiating as we anyway force 1/1
as output

9 years agoglvideomixer, glmixer: Add description and klass
Olivier Crête [Fri, 3 Jul 2015 16:17:42 +0000 (12:17 -0400)]
glvideomixer, glmixer: Add description and klass

9 years agogl: Don't leak pool if set_config failed
Nicolas Dufresne [Fri, 26 Jun 2015 19:34:35 +0000 (15:34 -0400)]
gl: Don't leak pool if set_config failed

9 years agovideoaggregator: simplifies and improves sink_get_caps.
Mathieu Duponchelle [Sat, 20 Jun 2015 11:36:27 +0000 (13:36 +0200)]
videoaggregator: simplifies and improves sink_get_caps.

The problem here was that after removing the formats and
all the things we could convert, we then intersected these
caps with the template caps.

Hence if a subclass offered permissive sink templates
(eg all the possible formats videoconvert handles), but only
one output format, then at negotiation time getcaps returned
caps with the format restricted to that format, even though
we do handle conversion.

https://bugzilla.gnome.org/show_bug.cgi?id=751255

9 years agogl: Add glviewconvert, glstereomix and glstereosplit elements
Jan Schmidt [Fri, 29 May 2015 16:29:04 +0000 (02:29 +1000)]
gl: Add glviewconvert, glstereomix and glstereosplit elements

Conversion elements for transforming multiview/stereoscopic video

https://bugzilla.gnome.org/show_bug.cgi?id=611157

9 years agovideoaggregator: Print some debug output if we change the timestamp offset
Sebastian Dröge [Mon, 15 Jun 2015 12:29:04 +0000 (14:29 +0200)]
videoaggregator: Print some debug output if we change the timestamp offset

9 years agovideoaggregator: Add some more debug output
Sebastian Dröge [Mon, 15 Jun 2015 12:25:43 +0000 (14:25 +0200)]
videoaggregator: Add some more debug output

9 years agovideoaggregator: Don't update the ts-offset before updating the actual configured...
Sebastian Dröge [Mon, 15 Jun 2015 12:19:05 +0000 (14:19 +0200)]
videoaggregator: Don't update the ts-offset before updating the actual configured caps

9 years agovideoaggregator: Remember if the last LATENCY query returned live or not and use...
Sebastian Dröge [Mon, 15 Jun 2015 12:18:39 +0000 (14:18 +0200)]
videoaggregator: Remember if the last LATENCY query returned live or not and use the in the QoS messages

9 years agocompositor: update zorder documentation.
Mathieu Duponchelle [Sun, 14 Jun 2015 21:20:38 +0000 (23:20 +0200)]
compositor: update zorder documentation.

It is not bound between 0 and 10000 anymore.

9 years agovideoaggregator: No need to artificially bound the zorder.
Mathieu Duponchelle [Sun, 14 Jun 2015 21:13:59 +0000 (23:13 +0200)]
videoaggregator: No need to artificially bound the zorder.

It is an unsigned integer so the upper bound is G_MAXUINT.

9 years agoglmixerbin: implement proper dynamic pad removal
Matthew Waters [Sat, 13 Jun 2015 08:43:04 +0000 (18:43 +1000)]
glmixerbin: implement proper dynamic pad removal

https://bugzilla.gnome.org/show_bug.cgi?id=750881

9 years agotests: Add test for the 'ignore-eos' compositor sink pad property
Nirbheek Chauhan [Sat, 6 Jun 2015 15:10:13 +0000 (20:40 +0530)]
tests: Add test for the 'ignore-eos' compositor sink pad property

When the 'ignore-eos' property is set on a pad, compositor will keep resending
the last buffer on the pad till the pad is unlinked. We count the buffers
received on appsink, and if it's more than the buffers sent by videotestsrc, the
test passes.

9 years agovideoaggregator: add "ignore-eos" property for input pads
Nirbheek Chauhan [Mon, 9 Feb 2015 19:19:35 +0000 (00:49 +0530)]
videoaggregator: add "ignore-eos" property for input pads

When set, it causes videoaggregator to repeatedly aggregate the last buffer on
an EOS pad instead of skipping it and outputting silence. This is useful, for
instance, while playing back files seamless one after the other, to avoid
videoaggregator ever outputting silence (the checkerboard pattern).

It is to be noted that if all the pads on videoaggregator have this property set
on them, the mixer will never forward EOS downstream for obvious reasons. Hence,
at least one pad with 'ignore-eos' set to FALSE must send EOS to the mixer
before it will be forwarded downstream.

https://bugzilla.gnome.org/show_bug.cgi?id=748946

9 years agogl: Use gst_object_ref_sink() for gl{filter,mixer,src}bin too
Sebastian Dröge [Thu, 11 Jun 2015 13:22:04 +0000 (15:22 +0200)]
gl: Use gst_object_ref_sink() for gl{filter,mixer,src}bin too

9 years agovideoaggregator: simplify aggregate returning
Thiago Santos [Wed, 6 May 2015 17:29:01 +0000 (14:29 -0300)]
videoaggregator: simplify aggregate returning

Rework special handling with goto/labels to only have one case
and otherwise just return normally.

9 years agovideoaggregator: refactor caps reconfigure to its own function
Thiago Santos [Wed, 6 May 2015 17:19:36 +0000 (14:19 -0300)]
videoaggregator: refactor caps reconfigure to its own function

Makes the aggregation code shorter and easier to read

9 years agovideoaggregator: fixing types in aggregate function
Thiago Santos [Wed, 6 May 2015 16:59:51 +0000 (13:59 -0300)]
videoaggregator: fixing types in aggregate function

Correctly use boolean and GstFlowReturn types in the function.

9 years agovideoaggregator: use macro to access aggregator src pad
Thiago Santos [Wed, 6 May 2015 16:08:11 +0000 (13:08 -0300)]
videoaggregator: use macro to access aggregator src pad

Makes code a bit more readable

9 years agoRelease 1.5.1
Sebastian Dröge [Sun, 7 Jun 2015 08:55:35 +0000 (10:55 +0200)]
Release 1.5.1

9 years agoFix a common typo: retreive -> retrieve
Thiago Santos [Fri, 5 Jun 2015 12:35:39 +0000 (09:35 -0300)]
Fix a common typo: retreive -> retrieve

Seems to have been copy pasted around a few places

9 years agovideoaggregator: Add class property to disable caps scaling
Jan Schmidt [Sun, 10 May 2015 08:55:16 +0000 (18:55 +1000)]
videoaggregator: Add class property to disable caps scaling

Add preserve_update_caps_result boolean on the class to allow
sub-classes to disable videoaggregator removing sizes and framerate
from the update_caps() return result.

9 years agovideoaggregator: Catch errors, and allow sub-class to return NULL from get_output_buf...
Jan Schmidt [Mon, 4 May 2015 08:17:21 +0000 (18:17 +1000)]
videoaggregator: Catch errors, and allow sub-class to return NULL from get_output_buffer()

A return value of GST_FLOW_OK with a NULL buffer from get_output_buffer()
means the sub-class doesn't want to produce an output buffer, so
skip it.

If gst_videoaggregator_do_aggregate() generates an error, make sure
to propagate it - don't just ignore and discard the error by
over-writing it with the gst_pad_push() result.

9 years agocompositor/glvideomixer: Don't calculate PAR/DAR with unset GstVideoInfos
Sebastian Dröge [Thu, 21 May 2015 13:24:48 +0000 (16:24 +0300)]
compositor/glvideomixer: Don't calculate PAR/DAR with unset GstVideoInfos

Otherwise we divide by zero.

9 years agocompositor: Fix double assignment
Sebastian Dröge [Thu, 21 May 2015 13:19:08 +0000 (16:19 +0300)]
compositor: Fix double assignment

9 years agocompositor/glvideomixer: fix up par handling
Matthew Waters [Wed, 20 May 2015 14:56:01 +0000 (00:56 +1000)]
compositor/glvideomixer: fix up par handling

We were using the wrong formula

https://bugzilla.gnome.org/show_bug.cgi?id=749634

9 years agoRevert "doc: Workaround gtkdoc issue"
Stefan Sauer [Mon, 18 May 2015 18:16:32 +0000 (20:16 +0200)]
Revert "doc: Workaround gtkdoc issue"

This reverts commit ff6c736fe08e01f4320c4b02e811a0b57cf97cc1.

This is fixed by the gtk-doc 1.23 release.

<para> cannot contain <refsect2>:
http://www.docbook.org/tdg/en/html/para.html
http://www.docbook.org/tdg/en/html/refsect2.html

9 years agodoc: Workaround gtkdoc issue
Nicolas Dufresne [Sun, 17 May 2015 03:38:14 +0000 (23:38 -0400)]
doc: Workaround gtkdoc issue

With gtkdoc 1.22, the XML generator fails when a itemizedlist is
followed by a refsect2. Workaround the issue by wrapping the
refsect2 into para.

9 years agogl: remove useless gl{En,Dis}able (GL_TEXTURE_*) calls
Matthew Waters [Thu, 14 May 2015 06:42:09 +0000 (16:42 +1000)]
gl: remove useless gl{En,Dis}able (GL_TEXTURE_*) calls

We are using shaders everywhere and so they are not needed

9 years agogl: element buffers are part of vao state
Matthew Waters [Fri, 1 May 2015 02:04:28 +0000 (12:04 +1000)]
gl: element buffers are part of vao state

Use them as such.  They are also required for GL3 core profile support
with glDrawElements on OS X.

9 years agocompositor: implement proper par handling
Matthew Waters [Thu, 14 May 2015 03:04:21 +0000 (13:04 +1000)]
compositor: implement proper par handling

We were previously failing on different input and output par

9 years agogl: don't deadlock on context creation failure
Matthew Waters [Wed, 13 May 2015 13:54:52 +0000 (23:54 +1000)]
gl: don't deadlock on context creation failure

https://bugzilla.gnome.org/show_bug.cgi?id=749284

9 years agoglmixerbin: Don't unref pad templates
Sebastian Dröge [Wed, 13 May 2015 12:42:15 +0000 (15:42 +0300)]
glmixerbin: Don't unref pad templates

Otherwise we unref the reference that is owned by the element class.

9 years agocompositor: fix rectangle obscure test to clamp against the output frame size
Matthew Waters [Wed, 13 May 2015 07:38:35 +0000 (17:38 +1000)]
compositor: fix rectangle obscure test to clamp against the output frame size

Rather than one of the input pad video info's.

The test checking this was not constraining the output frame size
to ensure that the out of frame stream was not being displayed.

9 years agoglvideomixer: implement par handling
Matthew Waters [Wed, 13 May 2015 07:11:55 +0000 (17:11 +1000)]
glvideomixer: implement par handling

We were previously ignoring it completely

9 years agoglvideomixer: don't upload the vertex data every frame
Matthew Waters [Wed, 13 May 2015 07:10:42 +0000 (17:10 +1000)]
glvideomixer: don't upload the vertex data every frame

Add the missing cache tracking statement.

9 years agoglmixer: Implement GstVideoAggregator::find_best_format()
Sebastian Dröge [Mon, 11 May 2015 12:54:52 +0000 (15:54 +0300)]
glmixer: Implement GstVideoAggregator::find_best_format()

Without this, we will fixate weird pixel-aspect-ratios like 1/2147483647. But
in the end, all the negotiation code in videoaggregator needs a big cleanup
and videoaggregator needs to get rid of the software-mixer specific things
everywhere.

9 years agoRevert "gl: readd glupload/download onto element pads"
Sebastian Dröge [Wed, 6 May 2015 13:46:49 +0000 (15:46 +0200)]
Revert "gl: readd glupload/download onto element pads"

This reverts commit 87d8270f302b03f63ce04f986d824892a2c131fd.

9 years agoRevert "Revert "glvideomixer: implement with glmixerbin""
Sebastian Dröge [Wed, 6 May 2015 13:43:32 +0000 (15:43 +0200)]
Revert "Revert "glvideomixer: implement with glmixerbin""

This reverts commit b4bd11f2f3a60224d188b27ab55b278077cb1217.

9 years agotests: Add a check for the new compositor pad-is-obscured optimization
Nirbheek Chauhan [Fri, 13 Mar 2015 00:40:52 +0000 (06:10 +0530)]
tests: Add a check for the new compositor pad-is-obscured optimization

We verify that all the buffers on an obscured sinkpad are skipped by overriding
the map() function in the GstVideoMeta of the buffers to set a variable when
called. We also test that the buffers do get mapped when they're not obscured.

Blame^WCredit for the GstVideoMeta map() idea goes to Tim.

https://bugzilla.gnome.org/show_bug.cgi?id=746147

9 years agocompositor: Only map the frame from a buffer if it will be used
Nirbheek Chauhan [Sat, 18 Apr 2015 09:40:00 +0000 (15:10 +0530)]
compositor: Only map the frame from a buffer if it will be used

It's a waste of resources to map it if it won't be converted
or used at all. Since we moved the frame mapping down, we need
to use the GST_VIDEO_INFO accessor macros now in the code above
that instead of the GST_VIDEO_FRAME accessor macros.

https://bugzilla.gnome.org/show_bug.cgi?id=746147

9 years agocompositor: use accessor macros for consistency
Nirbheek Chauhan [Sat, 18 Apr 2015 09:39:02 +0000 (15:09 +0530)]
compositor: use accessor macros for consistency

https://bugzilla.gnome.org/show_bug.cgi?id=746147

9 years agocompositor: Skip pads that are completely obscured by a higher zorder pad
Nirbheek Chauhan [Sat, 18 Apr 2015 09:39:02 +0000 (15:09 +0530)]
compositor: Skip pads that are completely obscured by a higher zorder pad

For each frame, compare the frame boundaries, check if the format contains an
alpha channel, check opacity, and skip the frame if it's going to be completely
overwritten by a higher zorder frame. The check is O(n^2), but that doesn't
matter here because the number of sinkpads is small.

More can be done to avoid needless drawing, but this covers the majority of
cases. See TODOs. Ideally, a reverse painter's algorithm should be used for
optimal drawing, but memcpy during compositing is small compared to the CPU used
for frame conversion on each pad.

https://bugzilla.gnome.org/show_bug.cgi?id=746147

9 years agogl: readd glupload/download onto element pads
Matthew Waters [Thu, 30 Apr 2015 01:15:40 +0000 (11:15 +1000)]
gl: readd glupload/download onto element pads

Allows insertion of gl elements into non-gl pipelines without converter
(upload/download) elements.

https://bugzilla.gnome.org/show_bug.cgi?id=743974

9 years agoRevert "glvideomixer: implement with glmixerbin"
Matthew Waters [Wed, 29 Apr 2015 12:55:00 +0000 (22:55 +1000)]
Revert "glvideomixer: implement with glmixerbin"

This reverts commit 0fb56738a14391f248aa0be8756adeaf978baa0c.

9 years agoglupload: provide the sink template caps that could be used
Matthew Waters [Tue, 28 Apr 2015 10:11:07 +0000 (20:11 +1000)]
glupload: provide the sink template caps that could be used

https://bugzilla.gnome.org/show_bug.cgi?id=746399

9 years agogl: unref display/other-context in the correct place
Matthew Waters [Mon, 27 Apr 2015 05:20:56 +0000 (15:20 +1000)]
gl: unref display/other-context in the correct place

Otherwise state changes from PLAYING->READY->PAUSED will cause there to
to be no display configured on the element.

https://bugzilla.gnome.org/show_bug.cgi?id=748405

9 years agoglmixer: Possible null pointer dereference
Vineeth T M [Wed, 22 Apr 2015 09:54:45 +0000 (18:54 +0900)]
glmixer: Possible null pointer dereference

While printing error message when context fails, error variable is not being used anymore
so it will lead to null pointer dereference

https://bugzilla.gnome.org/show_bug.cgi?id=748287

9 years agogldisplay: synchronize the searching and creation of GstGLContext's
Matthew Waters [Tue, 21 Apr 2015 02:19:46 +0000 (12:19 +1000)]
gldisplay: synchronize the searching and creation of GstGLContext's

Ootherwise we could end up with multiple elements in different chains
each creating a context.  Fixes context creation with glvideomixer.

9 years agoglmixer: fix caps leak in gst_gl_mixer_pad_sink_getcaps()
Guillaume Desmottes [Wed, 15 Apr 2015 13:22:37 +0000 (15:22 +0200)]
glmixer: fix caps leak in gst_gl_mixer_pad_sink_getcaps()

Caps refcounting was all wrong in this function. Rewrote it and add some
comments to make it clearer.

Fix caps leaks with the
validate.file.glvideomixer.simple.play_15s.synchronized scenario.

https://bugzilla.gnome.org/show_bug.cgi?id=747915

Signed-off-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>
9 years agovideoaggregator: fix caps leak when early returning
Guillaume Desmottes [Thu, 16 Apr 2015 12:17:04 +0000 (14:17 +0200)]
videoaggregator: fix caps leak when early returning

https://bugzilla.gnome.org/show_bug.cgi?id=747993

Signed-off-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>
9 years agoglmixer: unref owned caps when finalizing the mixer
Guillaume Desmottes [Thu, 16 Apr 2015 12:21:16 +0000 (14:21 +0200)]
glmixer: unref owned caps when finalizing the mixer

Fix a caps leak with the
validate.file.glvideomixer.simple.play_15s.synchronized scenario.

https://bugzilla.gnome.org/show_bug.cgi?id=747915

Signed-off-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>
9 years agoglmixer: pass the proper free function to frames and buffers array
Guillaume Desmottes [Wed, 15 Apr 2015 13:16:33 +0000 (15:16 +0200)]
glmixer: pass the proper free function to frames and buffers array

'array_buffers' contain borrowed GstBuffer and so shouldn't have a free
function. 'frames' is the one containing GstGLMixerFrameData and so should use
_free_glmixer_frame_data as free function.

Fix GstGLMixerFrameData leaks with the
validate.file.glvideomixer.simple.play_15s.synchronized scenario.

https://bugzilla.gnome.org/show_bug.cgi?id=747913

Signed-off-by: Guillaume Desmottes <guillaume.desmottes@collabora.co.uk>
9 years agogl: Remove some empty ::finalize() implementations
Sebastian Dröge [Fri, 17 Apr 2015 12:09:47 +0000 (14:09 +0200)]
gl: Remove some empty ::finalize() implementations

9 years agovideoaggregator: Remove broken timestamps-going-backwards check
Sebastian Dröge [Thu, 9 Apr 2015 02:29:40 +0000 (19:29 -0700)]
videoaggregator: Remove broken timestamps-going-backwards check

This would've also triggered if for some reason the segment was updated
in such a way that PTS went backwards, but the running time increased. Like
what happens when non-flushing seeks are done.

We're doing a proper buffer-from-the-past check a few lines below based on the
running time, which is the only time we should care about here.

9 years agoaggregator: Query latency on first incoming buffer.
Olivier Crête [Sat, 7 Mar 2015 02:12:52 +0000 (21:12 -0500)]
aggregator: Query latency on first incoming buffer.

And keep on querying upstream until we get a reply.

Also, the _get_latency_unlocked() method required being calld
with a private lock, so removed the _unlocked() variant from the API.
And it now returns GST_CLOCK_TIME_NONE when the element is not live as
we think that 0 upstream latency is possible.

https://bugzilla.gnome.org/show_bug.cgi?id=745768

9 years agoglvideomixer: Don't use context if not present
Edward Hervey [Tue, 31 Mar 2015 14:27:00 +0000 (16:27 +0200)]
glvideomixer: Don't use context if not present

Avoids assertions at runtime

9 years agovideoaggregator: Check if there's a previous buffer to 'keep'
Nirbheek Chauhan [Tue, 24 Mar 2015 13:34:26 +0000 (19:04 +0530)]
videoaggregator: Check if there's a previous buffer to 'keep'

If we want to keep a previous buffer but there's no previous buffer, we actually
need more data instead.

9 years agocompositor: Revert most of previous patch.
Jan Schmidt [Sun, 15 Mar 2015 16:53:33 +0000 (03:53 +1100)]
compositor: Revert most of previous patch.

The calculation doesn't produce the same results.
Keep just the change to divide alpha by 255 instead of 256,
for slightly better accuracy

9 years agocompositor: Fix blending functions
Jan Schmidt [Sun, 15 Mar 2015 13:22:14 +0000 (00:22 +1100)]
compositor: Fix blending functions

Correctly calculate alpha in a few places by dividing by 255,
not 256.

Fix the argb and bgra blending functions to avoid an off-by-one
error in the calculations, so painting with alpha = 0xff doesn't
ever bleed through from behind

9 years agoglsyncmeta: make context to wait and set sync explicit
Matthew Waters [Sat, 14 Mar 2015 16:25:29 +0000 (16:25 +0000)]
glsyncmeta: make context to wait and set sync explicit

otherwise we may wait on a sync object in same context by accident