Doing so involves retrieving the current viewport from OpenGL which as
with any glGet operation, is expensive.
This means that the various sinks need to reset the viewport on draw.
In the process, fix resizing on cocoa.
If we only ever make it to READY, transform_caps can create an
internal convert object that will never be freed by basetransform's
stop vmethod (PAUSED->READY).
The use of mediump as a specifier in GLSL shaders will have limited
resolution and when used as texture coordinates may become inaccurate
over texture sizes of 1024.
This reverts commit 69c3c31608.
All devices have the same name, they are duplicated with pulseaudio one
and the provided does not respond to HW being plugged/unplugged. I think
it's not ready for 1.16.
Binding the vertex array to 0 will unbind everything else already.
In the previous order older versions of the Intel GL driver caused
errors to be printed for every single call when disabling the vertex
attrib arrays after binding the vertex array to 0.
It might be the case that glgsinkbin would try to set a property to
its internal sink which doesn't exist in it, leading to a glib's
warning. For example, when playsink sets 'force-aspect-ratio' property
and glsinkbin has, as internal sink, appsink, which doesn't handle
that property.
The patch validates the incoming property to forward to internal sink
if it exists in the internal sink and both properties has the same
type.
Running the context query in _start and during the NULL->READY state transition
can fail because downstream elements might not be able to answer and thus the
source element would not be able to reuse downstream GLContext and GLDisplay.
This issue happened specifically when trying to use gltestsrc in playbin.
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
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
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
... instead of waiting for the first non-header buffer.
Also drop non-identification headers arriving after initialization or
before the identification header. We don't do anything with them and
they would just accumulate.
https://bugzilla.gnome.org/show_bug.cgi?id=796980
The argument 0x0 is interpreted by the x86 compiler as a 32-bit int, but
it is consumed as a 64-bit uint causing a segmentation fault. We need to
explicit cast it to guint64 in order for the va_list to be built correctly.
https://bugzilla.gnome.org/show_bug.cgi?id=797092
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
It makes sense to control it explicitly to allow us to enable it on
platforms that don't have hardware floating-point, and to allow people
to enable the 'vorbis' plugin without having to also provide the
Tremor dependency which is useless on most devices.
And clean up any old pending headers if we receive a new identification
header, or if we receive a new set of headers via caps.
Otherwise it might happen that we receive one or more header but not
all, and then afterwards all headers again, and libvorbis does not like
getting headers passed multiple times and would error out.
It only makes sense to pass the very latest headers to the decoder at
the time we can actually make use of them.
https://bugzilla.gnome.org/show_bug.cgi?id=796980
PangoCairo is thread-safe as long as the context and fontmap are not
shared between threads. Previously each subclass had its own context and
a class mutex for this reason, but apart from hurting performance this
was also not completely safe yet: the same fontmap might've been used by
different classes from different threads as the thread-default fontmap
(at time of class initialization) was used.
This effectively (but optionally) requires libjpeg-turbo which
ships with a .pc file and is what pretty much everyone these days
uses anyway for libjpeg, so shouldn't be a problem hopefully.
https://bugzilla.gnome.org/show_bug.cgi?id=796947
When enabled, phase-inversion slightly increase stereo quality, but
produce a stream that when downmixed to mono will present important
audio distortion. This patch disables this feature by default and
introduce a property that let user enable it if desired.
https://bugzilla.gnome.org/show_bug.cgi?id=791771
Matches the output from a similar glimagesink pipeline when
rotating from an upstream gltransformation passed through
the affine transformation meta with xpos/ypos being set.
https://bugzilla.gnome.org/show_bug.cgi?id=794401
e4bf9ed8f0 was not quite right and changed
the wrong thing. Intead we needed to change the multiplication order
and should have kept the previous to/from matrices as is done in this
patch.
The matrices were converting the wrong values with non-diagonal-only matrices.
e.g. a typical yflip matrix in [-1,1]^3 such as
1 0 0 0
0 -1 0 0
0 0 1 0
0 0 0 1
Would have actually required a matrix like this in [0,1]^3
1 0 0 0
0 -1 0 0
0 0 1 0
0 -2 0 1
Which is
1. not consistent with our multiplication convention and would require
transposing matrices or changing our multiplication order (from what is
generally used on opengl matrix guides/tutorials).
2. Produces incorrect values when input with actual vertices accounting for
the difference in multiplication order. e.g. some vertices multiplied by
the yflip matrix using vertex * yflip(== transpose(yflip) * vertex):
vertex: -> result: expected:
vec4(1,0,1,1) -> vec4(1,-2,1,1) vec4(1,1,1,1)
vec4(1,1,1,1) -> vec4(1,-3,1,1) vec4(1,0,1,1)
With the updated values, we now get the expected values.
Includes a test for this behaviour and the example above
A granule is a 64bit signed integer, shifting by 63 or more is
undefined and most likely an indication that the stream is
corrupted or invalid.
Detected by oss-fuzz
glcolorbalance is in the default GL path inside glimagesink,
so has been causing an possibly-unnecessary extra texture copy
on Android for a while now. If we're just doing passthrough,
we can support OES directly. If not, they'll be transformed
to 2D textures and colourbalanced.
This moves all the conversion related code to a single place, allows
less code-duplication inside compositor and makes the glmixer code less
awkward. It's also the same pattern as used by GstAudioAggregator.
The aggregated_frame is now called prepared_frame and passed to the
prepare_frame and cleanup_frame virtual methods directly. For the
currently queued buffer there is a method on the video aggregator pad
now.
It performs similarly to the existing alpha element however performs
calculations in floating point rather than with small (guint8) integers
so some differences are to be expected.
https://bugzilla.gnome.org/show_bug.cgi?id=794070
Previously we assumed that the texture ID is going to be valid even
after unmapping the frame, as it was immediately unmapped before even
being used. Now we only unmap once we're done with the texture.
