platform/upstream/gstreamer.git
9 years agoglwindow_x11: use parent default implementation
Julien Isorce [Wed, 27 May 2015 13:10:16 +0000 (14:10 +0100)]
glwindow_x11: use parent default implementation

9 years agoglcontext_glx: rename variable from window to context
Julien Isorce [Wed, 27 May 2015 10:16:32 +0000 (11:16 +0100)]
glcontext_glx: rename variable from window to context

9 years agoglwindow: initialize debug in class_init
Julien Isorce [Wed, 27 May 2015 10:11:55 +0000 (11:11 +0100)]
glwindow: initialize debug in class_init

Useful if one uses gst_gl_window_x11_new directly
instead of gst_gl_window_new.

9 years agoglcontext: initialize debug in class_init
Julien Isorce [Wed, 27 May 2015 10:07:38 +0000 (11:07 +0100)]
glcontext: initialize debug in class_init

Useful if one uses gst_gl_context_egl_new directly
instead of gst_gl_context_new.

9 years agoglcontext: add gst_gl_context_set_display helper
Julien Isorce [Wed, 27 May 2015 09:55:20 +0000 (10:55 +0100)]
glcontext: add gst_gl_context_set_display helper

9 years agoglcontext: add missing _CAST after G_TYPE_CHECK_CLASS
Julien Isorce [Wed, 27 May 2015 14:25:40 +0000 (15:25 +0100)]
glcontext: add missing _CAST after G_TYPE_CHECK_CLASS

9 years agoglwindow: add missing _CAST after G_TYPE_CHECK_CLASS
Julien Isorce [Wed, 27 May 2015 09:51:54 +0000 (10:51 +0100)]
glwindow: add missing _CAST after G_TYPE_CHECK_CLASS

9 years agoglwindow: provides some default implementations to factorize with all backends
Julien Isorce [Tue, 26 May 2015 17:38:39 +0000 (18:38 +0100)]
glwindow: provides some default implementations to factorize with all backends

9 years agoaggregator: Push EOS on error return.
Jan Schmidt [Wed, 27 May 2015 14:59:39 +0000 (00:59 +1000)]
aggregator: Push EOS on error return.

Before shutting down the srcpad task due to a
downstream error, push an EOS to give downstream
a chance to shut down somewhat cleanly.

9 years agompegtsmux: don't overwrite value
Luis de Bethencourt [Wed, 27 May 2015 12:12:57 +0000 (13:12 +0100)]
mpegtsmux: don't overwrite value

Value of res is reset to FALSE in each iteration of the while loop. We want to
conserve TRUE if any pad event succeeded until we arrive to done.

Also, buf is set to the value of *outbuf twice. Removing the first assignment
since the second one is outside of a conditional.

9 years agoopencv: 'for' loop initial declaration
Luis de Bethencourt [Wed, 27 May 2015 12:08:46 +0000 (13:08 +0100)]
opencv: 'for' loop initial declaration

'for' loop initial declarations are not allowed in C89, moving the declarations
to before the 'for' loops.

9 years agogleffects: 'for' loop initial declaration
Luis de Bethencourt [Wed, 27 May 2015 12:05:11 +0000 (13:05 +0100)]
gleffects: 'for' loop initial declaration

'for' loop initial declarations are not allowed in C89, moving the declarations
to before the 'for' loops.

9 years agoresindvd: remove unused value
Luis de Bethencourt [Wed, 27 May 2015 11:03:42 +0000 (12:03 +0100)]
resindvd: remove unused value

The data variable is incremented but never read again. Remove the unused value
assignment.

9 years agohlsdemux: remove more unused attributes
Thiago Santos [Wed, 27 May 2015 09:17:49 +0000 (06:17 -0300)]
hlsdemux: remove more unused attributes

Those are not used or only read

9 years agohlsdemux: remove unused attribute
Thiago Santos [Wed, 27 May 2015 08:35:30 +0000 (05:35 -0300)]
hlsdemux: remove unused attribute

It is never read for anything useful

9 years agogloverlay: fix a leak
Xavier Claessens [Tue, 26 May 2015 20:47:47 +0000 (16:47 -0400)]
gloverlay: fix a leak

https://bugzilla.gnome.org/show_bug.cgi?id=749846

9 years agogloverlay: properly handle errors while loading file
Xavier Claessens [Tue, 26 May 2015 19:10:28 +0000 (15:10 -0400)]
gloverlay: properly handle errors while loading file

Post an error on the bus if anything bad happens while reading
and parsing the image file.

https://bugzilla.gnome.org/show_bug.cgi?id=749846

9 years agogloverlay: remove unused type_file field
Xavier Claessens [Tue, 26 May 2015 19:04:34 +0000 (15:04 -0400)]
gloverlay: remove unused type_file field

https://bugzilla.gnome.org/show_bug.cgi?id=749846

9 years agogl: propagate return value from filter and filter_texture
Xavier Claessens [Tue, 26 May 2015 19:01:37 +0000 (15:01 -0400)]
gl: propagate return value from filter and filter_texture

https://bugzilla.gnome.org/show_bug.cgi?id=749846

9 years agodvbsuboverlay: remove unused assignment
Luis de Bethencourt [Tue, 26 May 2015 15:39:59 +0000 (16:39 +0100)]
dvbsuboverlay: remove unused assignment

buf is incremented just before returning, this new value is never used.
Removing unused assignment.

9 years agoadaptivedemux: 'for' loop initial declaration
Luis de Bethencourt [Tue, 26 May 2015 14:24:52 +0000 (15:24 +0100)]
adaptivedemux: 'for' loop initial declaration

'for' loop initial declarations are only allowed in C89, moving the declaration
to before the 'for' loop.

9 years agoexclusion: exception when set factor to 0
Vineeth T M [Tue, 26 May 2015 13:14:34 +0000 (14:14 +0100)]
exclusion: exception when set factor to 0

When factor property is set to 0, transform just returns.
Adjust the minimum value to 1.

https://bugzilla.gnome.org/show_bug.cgi?id=743907

9 years agogleffects: properly initialize the shaders across contexts implementing multiple...
Matthew Waters [Tue, 26 May 2015 02:47:40 +0000 (12:47 +1000)]
gleffects: properly initialize the shaders across contexts implementing multiple API's

9 years agogleffects_laplacian: fix shader compilation in gl3/gles2
Matthew Waters [Tue, 26 May 2015 02:44:47 +0000 (12:44 +1000)]
gleffects_laplacian: fix shader compilation in gl3/gles2

https://bugzilla.gnome.org/show_bug.cgi?id=748393

9 years agocurlsink: handle socket callback for active FTP connections as well
Patricia Muscalu [Mon, 25 May 2015 10:33:50 +0000 (12:33 +0200)]
curlsink: handle socket callback for active FTP connections as well

Since version 7.28.0, libcurl allows application to set
socket options for active FTP connections.

Bump libcurl requirement to version tested.

https://bugzilla.gnome.org/show_bug.cgi?id=749815

9 years agodashdemux: Fix segfault introduced by segment iteration refactoring (2105a310).
David Waring [Wed, 20 May 2015 11:08:38 +0000 (12:08 +0100)]
dashdemux: Fix segfault introduced by segment iteration refactoring (2105a310).

The functions to get the next fragment, next fragment timestamp and to advance
to the next fragment need to work differently when stream->segments is NULL.
Use logic similar to that introduced by commit 2105a310 to perform these
functions.

https://bugzilla.gnome.org/show_bug.cgi?id=749684

9 years agodashdemux: use signed integet to represent segment index
Thiago Santos [Tue, 19 May 2015 01:33:50 +0000 (22:33 -0300)]
dashdemux: use signed integet to represent segment index

We need to be able to represent -1 to check when we are EOS
on reverse playback and unsigned integer won't allow that.

CID #1298193
CID #1298194
CID #1298195

https://bugzilla.gnome.org/show_bug.cgi?id=749534

9 years agodashdemux: fix outdated example launch line
Suhwang Kim [Mon, 25 May 2015 01:30:55 +0000 (10:30 +0900)]
dashdemux: fix outdated example launch line

https://bugzilla.gnome.org/show_bug.cgi?id=749809

9 years agogl/wayland: remove dead event source code
Matthew Waters [Fri, 22 May 2015 03:12:09 +0000 (13:12 +1000)]
gl/wayland: remove dead event source code

9 years agogl/wayland: add GstGLDisplayWayland
Matthew Waters [Thu, 21 May 2015 07:48:31 +0000 (17:48 +1000)]
gl/wayland: add GstGLDisplayWayland

Simple implementation split from GstGLWindowWayland

Can now have multiple glimagesink elements all displaying output
linked via GL or otherwise (barring GL platform limitations).

The intel driver is racy and can crash setting up the two glimagesink contexts.

e.g.
videotestsrc ! tee name=t ! queue ! glupload ! glimagesinkelement
  t. ! queue ! gleffects_blur ! glimagesinkelement

videotestsrc ! glupload ! glfiltercube ! tee name=t ! queue ! glimagesinkelement
  t. ! queue ! gleffects_blur ! glimagesinkelement

9 years agogltransformation: fix DrawElements call for element array buffers
Matthew Waters [Mon, 25 May 2015 07:27:58 +0000 (17:27 +1000)]
gltransformation: fix DrawElements call for element array buffers

https://bugzilla.gnome.org/show_bug.cgi?id=749734

9 years agogl/calayer: provide the exact GL api version for the wrapped context
Matthew Waters [Fri, 22 May 2015 15:00:18 +0000 (01:00 +1000)]
gl/calayer: provide the exact GL api version for the wrapped context

Otherwise we could end up being mistaken for the diference between a
gl3 and a gl2 context resulting in a failure getting the list of
extensions from the wrapped context due to the difference between
glGetString and glGetStringi for the GL_EXTENSIONS token.

https://bugzilla.gnome.org/show_bug.cgi?id=749728

9 years agodashdemux: remove unnecessary semicolon
Jimmy Ohn [Thu, 21 May 2015 17:40:37 +0000 (02:40 +0900)]
dashdemux: remove unnecessary semicolon

https://bugzilla.gnome.org/show_bug.cgi?id=749691

9 years agogl: win32: do not call SetParent in release_parent_win_id()
Xavier Claessens [Thu, 21 May 2015 19:30:34 +0000 (15:30 -0400)]
gl: win32: do not call SetParent in release_parent_win_id()

When called from gst_gl_window_win32_close(), internal window
could not exist, and if it does it's going to be destroyed just
after that anyway. Also it causes window_proc() to be called
and crash because it gets a NULL context.

When called from gst_gl_window_win32_set_window_handle() we are
going to set another parent anyway, and it's probably better to
reparent directly instead of passing by a NULL parent which could
cause the internal window to popup briefly.

https://bugzilla.gnome.org/show_bug.cgi?id=749601

9 years agogl: win32: remove unused code
Xavier Claessens [Thu, 21 May 2015 19:30:00 +0000 (15:30 -0400)]
gl: win32: remove unused code

https://bugzilla.gnome.org/show_bug.cgi?id=749601

9 years agogl: win32: fix crash when finalizing GstGLContext
Xavier Claessens [Wed, 20 May 2015 21:09:21 +0000 (17:09 -0400)]
gl: win32: fix crash when finalizing GstGLContext

gst_gl_context_finalize() is calling gst_gl_window_win32_quit()
which was posting a message. But then window_proc takes window's
context and get a NULL.

Now that we've got a GMainLoop we can do like other backends and
simply call g_main_loop_quit().

This also remove duplicated code to release the parent window and
potential crash there because parent_proc could be NULL if we never
created the internal window. That could happen for example if setting
state to READY then setting a window_handle, and go back to NULL state.

https://bugzilla.gnome.org/show_bug.cgi?id=749601

9 years agogl: win32: Fix leaked GstGLContext
Xavier Claessens [Wed, 20 May 2015 21:06:42 +0000 (17:06 -0400)]
gl: win32: Fix leaked GstGLContext

https://bugzilla.gnome.org/show_bug.cgi?id=749601

9 years agogl: win32: set the parent window when creating internal window
Xavier Claessens [Wed, 20 May 2015 19:30:49 +0000 (15:30 -0400)]
gl: win32: set the parent window when creating internal window

When _set_window_handle() was called in READY state, it wasn't
set to the internal window created later.

https://bugzilla.gnome.org/show_bug.cgi?id=749601

9 years agogl: win32: use a GMainContext to dispatch win32 messages
Xavier Claessens [Wed, 20 May 2015 19:29:50 +0000 (15:29 -0400)]
gl: win32: use a GMainContext to dispatch win32 messages

gst_gl_window_win32_send_message_async() could be called before the
internal window is created so we cannot use PostMessage there.

x11 and wayland backends both create a custom GSource for this,
so there is no reason to not do that for win32.

https://bugzilla.gnome.org/show_bug.cgi?id=749601

9 years agoglupload: Don't leak all memory uploaded via raw uploads.
Jan Schmidt [Thu, 21 May 2015 14:27:36 +0000 (00:27 +1000)]
glupload: Don't leak all memory uploaded via raw uploads.

Remove an extra ref missed when switching over to not
reusing output textures

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 agoglcontext: require a shader version to initialize.
Matthew Waters [Thu, 21 May 2015 05:05:33 +0000 (15:05 +1000)]
glcontext: require a shader version to initialize.

It's very near pointless to try our GL implementation without shaders.

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

9 years agoFix flags order in GL examples for uninstalled build
Jan Schmidt [Tue, 19 May 2015 16:38:53 +0000 (02:38 +1000)]
Fix flags order in GL examples for uninstalled build

9 years agowasapi: Fix flags order for uninstalled build
Jan Schmidt [Tue, 19 May 2015 15:57:52 +0000 (01:57 +1000)]
wasapi: Fix flags order for uninstalled build

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 agoh264parse: Consider SEI NALU as "HEADER" packets
Edward Hervey [Fri, 17 Apr 2015 13:36:49 +0000 (15:36 +0200)]
h264parse: Consider SEI NALU as "HEADER" packets

Like SPS/PPS they do contain information which will be needed to
decode the following data (as per definition of the flag)

Also ensures that the series of SPS/PPS/SEI NALU before a keyframe
can be considered as one contiguous header

9 years agompegtsmux: Carry over GST_BUFFER_FLAG_HEADER
Edward Hervey [Fri, 17 Apr 2015 13:35:43 +0000 (15:35 +0200)]
mpegtsmux: Carry over GST_BUFFER_FLAG_HEADER

In the same way we do it for the DELTA_UNIT flag

This allows downstream elements to know whether a given mpeg-ts
packet contains a corresponding HEADER elementary unit

9 years agox265enc: Only copy VPS/SPS/PPS NAL units in header buffer
Thijs Vermeir [Wed, 20 May 2015 10:11:31 +0000 (12:11 +0200)]
x265enc: Only copy VPS/SPS/PPS NAL units in header buffer

Previously the VPS unit was detected and all next packets where copied
into the header buffer assuming only SPS and PPS would follow.  This is
not always true, also other types of NAL units follow the VPS unit and
where copied to the header buffer. Now the VPS/SPS/PPS are explicitely
detected and copied in the header buffer.

9 years agox265enc: Fix tune parameter mismatch
Thijs Vermeir [Wed, 20 May 2015 09:27:25 +0000 (11:27 +0200)]
x265enc: Fix tune parameter mismatch

There was a mismatch between the tune parameter in x265 and the enum
used in this element. The value in the enum is the x265 tune
parameter + 1.

9 years agoamcaudiodec: Add an output adapter for chunking the output into codec frames
Sebastian Dröge [Tue, 19 May 2015 15:02:30 +0000 (18:02 +0300)]
amcaudiodec: Add an output adapter for chunking the output into codec frames

Otherwise the base class will be confused.
See https://bugzilla.gnome.org/show_bug.cgi?id=685730

9 years agogl: fix crash on windows with intel driver
Xavier Claessens [Tue, 19 May 2015 14:43:28 +0000 (10:43 -0400)]
gl: fix crash on windows with intel driver

https://bugzilla.gnome.org/show_bug.cgi?id=749430

9 years agocaopengllayersink: static const the indices array
Matthew Waters [Tue, 19 May 2015 14:50:37 +0000 (00:50 +1000)]
caopengllayersink: static const the indices array

the contents will never change anyway.

9 years agocaopengllayersink: correctly use the sync meta across multiple contexts
Matthew Waters [Tue, 19 May 2015 14:28:52 +0000 (00:28 +1000)]
caopengllayersink: correctly use the sync meta across multiple contexts

1. Set the sync point after the (possible) upload has occured
2. Wait in the correct GL context (the draw context)

Note: We don't add the GL sync meta to the input buffer as it's not
writable and a copy would be expensive.

Similar to the change with the same name for glimagesink

9 years agocaopengllayersink: update for GL3 element array buffer usage
Matthew Waters [Tue, 19 May 2015 14:28:07 +0000 (00:28 +1000)]
caopengllayersink: update for GL3 element array buffer usage

fixes blank output

9 years agoglimagesink: correctly use the sync meta across multiple contexts
Matthew Waters [Tue, 19 May 2015 06:22:00 +0000 (16:22 +1000)]
glimagesink: correctly use the sync meta across multiple contexts

1. Set the sync point after the (possible) upload has occured
2. Wait in the correct GL context (the draw context)

Note: We don't add the GL sync meta to the input buffer as it's not
writable and a copy would be expensive.

9 years agogl/calayer: provide a current wrapped GstGLContext for callbacks
Matthew Waters [Tue, 19 May 2015 06:19:41 +0000 (16:19 +1000)]
gl/calayer: provide a current wrapped GstGLContext for callbacks

So that the draw and resize callbacks can get the current GstGLContext.

9 years agoglsyncmeta: ensure that the Fence is going to be executed
Matthew Waters [Tue, 19 May 2015 06:14:39 +0000 (16:14 +1000)]
glsyncmeta: ensure that the Fence is going to be executed

Otherwise it could stay client side without being submitted to the GL
server resulting in another context waiting on a Fence that will never
become signalled causing a deadlock.

9 years agoglupload: don't reuse output textures for raw data upload
Matthew Waters [Tue, 19 May 2015 01:47:23 +0000 (11:47 +1000)]
glupload: don't reuse output textures for raw data upload

Causes stale textures to be used further down the chain.

9 years agoglframebuffer: don't clear the framebuffer
Matthew Waters [Tue, 19 May 2015 01:11:02 +0000 (11:11 +1000)]
glframebuffer: don't clear the framebuffer

Breaks attempting to blend with the destination buffer and should be done
explicitly by the callback anyway.

9 years agoglbasefilter: track gl_start/stop correctly
Matthew Waters [Tue, 19 May 2015 01:09:54 +0000 (11:09 +1000)]
glbasefilter: track gl_start/stop correctly

Don't start multiple times without calling gl_stop.

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 agohlsdemux: Simplify logic in process_manifest
Jimmy Ohn [Wed, 13 May 2015 17:11:50 +0000 (02:11 +0900)]
hlsdemux: Simplify logic in process_manifest

Simplify logic in process_manifest and remove a TODO item.

https://bugzilla.gnome.org/show_bug.cgi?id=749328

9 years agocvsmooth: rename properties
Vineeth T M [Mon, 18 May 2015 05:12:56 +0000 (14:12 +0900)]
cvsmooth: rename properties

The properties are named as param1 to param4, which makes very little sense
hence renamed the properties according to what it is doing.

https://bugzilla.gnome.org/show_bug.cgi?id=749523

9 years agocvsmooth: Wrong assignment of prop leads to crash
Vineeth T M [Mon, 18 May 2015 05:01:59 +0000 (14:01 +0900)]
cvsmooth: Wrong assignment of prop leads to crash

when setting param2 property, it is wrongly being assigned to param1.
This leads to wrong behavior and a crash when param2 is set as 0.

https://bugzilla.gnome.org/show_bug.cgi?id=749523

9 years agopyramidsegment: wrong value of level property
Vineeth T M [Mon, 18 May 2015 05:30:19 +0000 (14:30 +0900)]
pyramidsegment: wrong value of level property

The property level has a minimum value of 0. But when we set the level as 0,
it gets an assertion error. The function icvPyrSegmentation8uC3R returns false
if level is set as 0, since the minimum level cant be 0 and thus results in error.
Hence changing the minimum value to 1.

https://bugzilla.gnome.org/show_bug.cgi?id=749525

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 agotsdemux: fix buffer timestamp not being in stream time
Vincent Penquerc'h [Wed, 6 May 2015 14:17:24 +0000 (15:17 +0100)]
tsdemux: fix buffer timestamp not being in stream time

Timestamps should start at the segment start, rather than 0, so
we need to not subtract the first timestamp. This makes the sink
correctly account for running time when switching PMTs where a
stream starts not quite at zero, causing timing offsets that can
become noticeable and causing dropped frames after a few times.

9 years agotsdemux: accumulate previous segment base time
Vincent Penquerc'h [Wed, 6 May 2015 14:16:37 +0000 (15:16 +0100)]
tsdemux: accumulate previous segment base time

9 years agotsdemux: fix refcounting when applying a new PMT
Vincent Penquerc'h [Fri, 24 Apr 2015 11:56:02 +0000 (12:56 +0100)]
tsdemux: fix refcounting when applying a new PMT

A new program object is created to replace an existing one
in the programs hash table, so its refcount needs to match.

With the default of 0 refcount on creation, the next PAT
change will cause that refcount to be both incremented and
decremented (assuming the new PAT references that stream too),
which will cause the program to be destroyed.

https://bugzilla.gnome.org/show_bug.cgi?id=748412

9 years agodtls: Fix memory leak
Jose Antonio Santos Cadenas [Thu, 14 May 2015 14:06:55 +0000 (16:06 +0200)]
dtls: Fix memory leak

Keys were not correctly released when it was get as a property
nor when a second key was received

https://bugzilla.gnome.org/show_bug.cgi?id=749380

9 years agoglimagesink: free the vertex buffer when done
Matthew Waters [Thu, 14 May 2015 11:21:01 +0000 (21:21 +1000)]
glimagesink: free the vertex buffer when done

fixes a memory leak

9 years agoglcontext/cocoa: implement GL3 core context selection
Matthew Waters [Thu, 22 Jan 2015 07:00:36 +0000 (18:00 +1100)]
glcontext/cocoa: implement GL3 core context selection

9 years agotests/gl: fix typo
Matthew Waters [Thu, 14 May 2015 08:35:35 +0000 (18:35 +1000)]
tests/gl: fix typo

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 agoglimagesink: cleanup properties
Matthew Waters [Thu, 14 May 2015 05:10:59 +0000 (15:10 +1000)]
glimagesink: cleanup properties

remove unused "display-name"
ensure defaults between the bin/element are the same

9 years agogleffects: only try the GL2 shader init path if we have a GL2 context
Matthew Waters [Thu, 14 May 2015 04:56:30 +0000 (14:56 +1000)]
gleffects: only try the GL2 shader init path if we have a GL2 context

9 years agoglshader: attempt to detect the gles2 inside opengl3 case
Matthew Waters [Tue, 12 May 2015 12:04:26 +0000 (22:04 +1000)]
glshader: attempt to detect the gles2 inside opengl3 case

This is necessary to use gles2 shaders in a GL 3 core context on
OS X which fails without a proper #version being set on the shaders.

9 years agoglshader: remove references to gl3 specific shaders
Matthew Waters [Tue, 12 May 2015 07:55:43 +0000 (17:55 +1000)]
glshader: remove references to gl3 specific shaders

We rely specifically on gles2 shaders being supported by the GL
implementation with GL3 core profile.

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 agoglmemory: properly detect (hopefully) the correct ext_rg/arb_rg variant
Matthew Waters [Thu, 7 May 2015 05:25:10 +0000 (12:25 +0700)]
glmemory: properly detect (hopefully) the correct ext_rg/arb_rg variant

GL_EXT_texture_rg is only valid for GLES2. GLES3 uses similar wording to
the GL_ARB_texture_rg which requires a sized internal format that the
GL_EXT_texture_rg does not require.

https://www.khronos.org/registry/gles/extensions/EXT/EXT_texture_rg.txt
https://www.opengl.org/registry/specs/ARB/texture_rg.txt

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 agodashdemux: avoid busy-looping when waiting for new fragment
Thiago Santos [Tue, 12 May 2015 16:41:19 +0000 (13:41 -0300)]
dashdemux: avoid busy-looping when waiting for new fragment

When all fragments have already been downloaded on a live stream
dashdemux would busy loop as the default implementation of
has_next_fragment would return TRUE. Implement it to correctly
signal if adaptivedemux should wait for the manifest update before
trying to get new fragments.

9 years agodashdemux: Improve live stream manifest update repositioning
Thiago Santos [Mon, 11 May 2015 17:19:20 +0000 (14:19 -0300)]
dashdemux: Improve live stream manifest update repositioning

When updating the manifest the timestamps on it might have changed a little
due to rounding and timescale conversions. If the change makes the timestamp
of the current segment to go up it makes dashdemux reposition to the previous
one causing one extra unnecessary download.

So when repositioning add an extra 10 microseconds to cover for that rounding
issues and increase the chance of falling in the same segment.

Additionally, also improve the time used when the client is already after the
last segment. Instead of using the last segment starting timestamp use the
final timestamp to make it reposition to the next one and not to the one that
has already been downloaded.

9 years agodashdemux: add more protection when acessing gptrarray entries
Thiago Santos [Mon, 11 May 2015 15:30:03 +0000 (12:30 -0300)]
dashdemux: add more protection when acessing gptrarray entries

The glib structure doesn't do range checking so we have to do it
ourselves.

Also adds some more debugging messages

9 years agodashdemux: remove unused functions
Thiago Santos [Fri, 8 May 2015 19:58:36 +0000 (16:58 -0300)]
dashdemux: remove unused functions

These functions of directly getting and setting segment indexes
are no longer useful as now we need 2 indexes: repeat and segment
index.

The only operations needed are advance_segment, going back to the
first one or seeking for a timestamp.

9 years agodashdemux: reuse seeking function to reduce repeated code
Thiago Santos [Fri, 8 May 2015 19:49:02 +0000 (16:49 -0300)]
dashdemux: reuse seeking function to reduce repeated code

Instead of writing a seek routine, just use the mpdparser function.

Also remove function that is not needed anymore

9 years agodashdemux: refactor segment iteration for better performance
Thiago Santos [Fri, 8 May 2015 19:31:44 +0000 (16:31 -0300)]
dashdemux: refactor segment iteration for better performance

Segments are now stored with their repeat counts instead of spanding
them to multiple segments. This caused advancing to the next segment
using a single index to have to iterate over the whole list every time.

This commit addresses this by storing both the segment index as well
as the repeat index and makes advancing to next segment just an
increment of the repeat or the segment index.

9 years agodashdemux: improve mpd parsing for repeated segment entries
Thiago Santos [Thu, 7 May 2015 19:30:48 +0000 (16:30 -0300)]
dashdemux: improve mpd parsing for repeated segment entries

Use a single segment to represent it internally to avoid using too
much memory. This has the drawback of issuing a linear search to
find the correct segment to play but this can be fixed by using
binary searches or caching the current position and just looking
for the next one.

https://bugzilla.gnome.org/show_bug.cgi?id=748369

9 years agoadaptivedemux: drop debug message that was wrong
Thiago Santos [Tue, 12 May 2015 15:55:42 +0000 (12:55 -0300)]
adaptivedemux: drop debug message that was wrong

EOS doesn't necessarily mean that no fragment was downloaded

9 years agoadaptivedemux: run gst-indent
Thiago Santos [Tue, 12 May 2015 15:55:17 +0000 (12:55 -0300)]
adaptivedemux: run gst-indent

To allow commiting on this file cleanly

9 years agompegtsmux: the parent is provided in the function
Thiago Santos [Mon, 11 May 2015 14:56:52 +0000 (11:56 -0300)]
mpegtsmux: the parent is provided in the function

No need to get it again

9 years agodtlsenc: Fix memory leak releasing connection_id
Jose Antonio Santos Cadenas [Wed, 13 May 2015 13:59:48 +0000 (15:59 +0200)]
dtlsenc: Fix memory leak releasing connection_id

https://bugzilla.gnome.org/show_bug.cgi?id=749318

9 years agodtlsenc: Fix memory leak while setting connection-id
Jose Antonio Santos Cadenas [Wed, 13 May 2015 13:33:49 +0000 (15:33 +0200)]
dtlsenc: Fix memory leak while setting connection-id

https://bugzilla.gnome.org/show_bug.cgi?id=749318

9 years agodtlsdec: Fix memory leak, release previous pem
Jose Antonio Santos Cadenas [Wed, 13 May 2015 15:15:57 +0000 (17:15 +0200)]
dtlsdec: Fix memory leak, release previous pem

https://bugzilla.gnome.org/show_bug.cgi?id=749322

9 years agodtlsdec: Fix memory leak on dispose
Jose Antonio Santos Cadenas [Wed, 13 May 2015 14:06:52 +0000 (16:06 +0200)]
dtlsdec: Fix memory leak on dispose

Parent dispose function was not called

https://bugzilla.gnome.org/show_bug.cgi?id=749322

9 years agodtlsconnection: Fix memory leak while setting closure
Jose Antonio Santos Cadenas [Wed, 13 May 2015 15:19:25 +0000 (17:19 +0200)]
dtlsconnection: Fix memory leak while setting closure

https://bugzilla.gnome.org/show_bug.cgi?id=749325

9 years agofix faad2 version check
Wim Taymans [Wed, 13 May 2015 14:23:26 +0000 (16:23 +0200)]
fix faad2 version check

On fedora 22, the output of cpp inserts extra debug comments, which
makes our regexp for the faad2 version check fail. This in turn causes
it to compile with the wrong arguments passed which then causes stack
corruption and crashes.

Fix this by only checking for the version (which should be by itself on
a single line). This is potentially less safe, it might be possible that
a similar string would appear in a later version in the header file.

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=748571

9 years agoglcontext: propagate the version/extension failure upwards
Matthew Waters [Wed, 13 May 2015 13:55:28 +0000 (23:55 +1000)]
glcontext: propagate the version/extension failure upwards

Otherwise the calling code has no idea that the context failed to
be created fully.

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

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