All UI elements will follow Single-Threaded Apartments (STA) model.
As a result, we should access them from dedicated UI thread.
Due to the nature of the threading model, ANGLE will wait the UI
thread while closing internal window/swapchain objects.
A problem here is that when destroying GstGLWindow from the UI thread,
it will wait GstGLContext's internal thread. Meanwhile, the GstGLContext's
internal thread will be blocked because ANGLE wants to access the UI thread.
That will cause a deadlock or exceptions.
In short, application should not try to call
gst_element_set_state(pipeline, GST_STATE_NULL) from a UI thread.
That's a limitation of current implementation.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/745>
This patch was taken from #629#note_178766, the comment made
at the time was:
The root issue is a mismatch between the initialization of render_rect
in GstGLWindowX11Private and what's expected in the draw_cb function.
Because render_rect is not explicitly initialized to a width and height
of -1 (unlike gstglwindow_wayland_egl.c which does initialize to -1),
the less-than check for explicitly-set render_rect at gstglwindow_x11.c:453-454
always fails, even when the parent_win has been set and the render rectangle
has never been set.
Maybe this came from copying the similar check in the wayland code? Regardless,
I think the correct inequality should be '<= 0' (on both lines).
Alternatively initialization could be changed, but other sinks, e.g.
xvimagesink don't appear to use -1 to mean "unset" render_rect this way.
The issue can be reproduced by running the example in
tests/examples/gl/qt/videooverlay/ on X11, and resizing the output
window
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/701>
This means we cannot access [view layer] or view.bounds from the OpenGL
thread. This also means that we need to call the main thread when
setting the window handle. However, we cannot perform that
synchronously as that may deadlock with the application performing the
set_window_handle() call.
We need to defer the actual update and run it asynchronously and wait
for the window handle update internally at each point it is needed.
Fixes: https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/issues/372
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/681>
The gir generator wrongly assume that the vfunc
GstGLFilterClass.filter() and the method gst_gl_filter_filter_texture()
are related. As a result it complains about not matching argument names.
Workaround this by naming both of their arguments input and output.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/664>
If there are two elements and threads attempting to query each other for
an OpenGL context. The locking may result in a deadlock.
We need to unlock each element's context_lock when querying another
element for the OpenGL context in order to allow any other element to
take the lock when the other element is querying for an OpenGL context.
Part-of: <https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/-/merge_requests/642>
Because the color value is stored in MSB, so we can reuse the
Y210 code for P012_LE / P012_BE
Sample pipeline:
gst-launch-1.0 videotestsrc ! video/x-raw,format=Y212_LE ! glimagesink
gst_gl_window_quit() will attempt to send a message but will be called
from GstGLContext's finalize handler and so the weak ref that backs
gst_gl_window_get_context will return NULL as it has already been
cleared. We need that context in send_message_async to decide whether
to run the provided callback immediately or queue in GCD
Without this fix, it is possible that outbuf is not initialized, which
will result in segfault when call gst_buffer_replace (&outbuf, NULL). In
addition, the patch fixes potential memory leak in restart path.
The segfault can be reproduced by the pipeline below:
GST_GL_PLATFORM=egl \
gst-launch-1.0 videotestsrc ! msdkh265enc ! msdkh265dec ! \
'video/x-raw(memory:DMABuf)' ! glimagesink
In the situation that the direct dmabuf path is chosen, but with an
unsupported texture format, this causes accept to fail rather than
continue and fail at the upload stage. It is also possibly necessary to
reconfigure after falling back from direct to non-direct dmabuf upload
paths.
This allows us to remove races when setting the wl_queue on wayland
objects with wl_proxy_set_queue() as each created object is created with
the queue already set.
We can also move all our initilization code into the window as we
can retrieve all wayland objects from each window instance. This
removes a possible race when integrating with external API's as we would
always attempt to immediately retrieve a small set of wayland objects.
That is no longer the case with the objects from each window instance.
Automatic negotiation of texture-target=external-oes does not work
without separating the external-oes support out of the DirectDmabuf
uploader into a separate DirectDmabufExternal uploader.
gst_gl_upload_transform_caps() is missing a NULL pointer check in case
the current upload method's transform_caps returns a NULL pointer. In
the following loop over all upload methods, NULL pointer return values
are already handled correctly.
Some drivers support directly importing DMA buffers in some formats into
external-oes textures only, for example because the hardware contains
native YUV samplers.
Note that in these cases colorimetry can only be passed as hints and
there is no feedback whether the driver supports the required YUV
encoding matrix and quantization range.
Allow creating EGL images from DMA buffers in formats that the driver
only supports for the external-oes texture target.
Pass the intended texture target to gst_egl_image_from_dmabuf_direct so
that _gst_egl_image_check_dmabuf_direct can decide whether to create an
EGL image for a format that can only be targeted at external-oes
textures by the driver. Allow creating GstGLMemoryEGL objects from these
DMA buffers.
ANGLE_surface_d3d_render_to_back_buffer extension is only available
with Microsoft fork of ANGLE. Note that Microsoft's ANGLE repository
has been deprecated.
Posting any message to parent seems to be pointless. That might break
parent window.
Regardless of the posting, parent window can catch mouse event
and also any keyboard events will be handled by parent window by default.
Fixes: https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/634
There was a typo in the extension name which resulted in the modifiers
to never be set when doing DMABuf import. That triggered the modifiers
lookup in Intel driver, which was in fact hiding bugs in the gldownload
to glupload path when doing DMABuf.
Note, this changes breaks pipeline the following pipeline on Intel and
some other drivers:
gltestsrc ! gldownload ! video/x-raw\(memory:DMABuf\) ! glimagsink
A fix for this was added to Mesa recently:
Related to https://gitlab.freedesktop.org/mesa/mesa/merge_requests/1338
Fixes 5d0e191710
We don't support modififers and that would result in bad image being
displayed. Note that this was fixes recently in Mesa MR 1138, prior to
that, the reported modifier is always 0, which makes this change a
no-op.
Fixes#441
Related to https://gitlab.freedesktop.org/mesa/mesa/merge_requests/1338
It's either this or replacing all the object lock usage in gldisplay
with a recursive mutex which is not backwards compatible
The failure case is effectively:
1. The user has locked the display object lock
2. a glcontext loses it's last ref and attempts to quit the window
3. gst_gl_window_quit() attempts to remove the window from the display
4. gst_gl_display_remove_window attempts to take the display object lock
The only concern with changing the locking for the window list in the
display is that gst_gl_display_create_window() has documentation requiring
the object lock to be held which must continue to work correctly.
Returning a transfer none value for a value checked by a lock is not
thread safe as the reference could disappear before the caller can take
its reference.
Y210 is a 10-bit YUY2, so we may re-use the YUY2 shaders but gl format
is set to RG16
Sample pipeline:
gst-launch-1.0 videotestsrc ! video/x-raw,format=Y210 ! glimagesink
NV16/NV61 is basically the same as NV12/NV21 with a higher chroma resolution.
Since only the size of the UV plane/texture is different, the same shaders are used as for NV12/NV21.
This is done by reusing `gst_gl_memory_setup_buffer` avoiding to
duplicate code.
Without a VideoMeta, mapping those buffers lead to GstBuffer mapping the
buffer in system memory even when specifying the GL flags (through the
buffer merging mechanism) making the result totally broken.
We are using ARC to cleanup after ourselves.
../gst-libs/gst/gl/cocoa/gstglwindow_cocoa.m:159:20: error: unused variable 'queue' [-Werror,-Wunused-variable]
dispatch_queue_t queue = (__bridge_transfer dispatch_queue_t) window->priv->gl_queue;
^
This is needed for using GstGL with ANGLE as the GLES implementation
in Universal Windows Platform apps that use the Windows Runtime
(WinRT) instead of Win32, which is deprecated and not allowed in
Windows Store apps.
This has been tested with Servo on the Microsoft HoloLens 2, and seems
to work quite well.
By passing NULL to `g_signal_new` instead of a marshaller, GLib will
actually internally optimize the signal (if the marshaller is available
in GLib itself) by also setting the valist marshaller. This makes the
signal emission a bit more performant than the regular marshalling,
which still needs to box into `GValue` and call libffi in case of a
generic marshaller.
Note that for custom marshallers, one would use
`g_signal_set_va_marshaller()` with the valist marshaller instead.
Simple addition for supporting EXT_platform_device typed display.
It's a kind of special display type (part of EGL specification)
which has no window at all.
To use EGLDevice explicitly, set environment "GST_GL_WINDOW=egl-device"
See also https://www.khronos.org/registry/EGL/extensions/EXT/EGL_EXT_platform_device.txt
The gltestsrc element was refactored to inherit from this base class which
handles the GL context. The sub-class only needs to implement the gl_start,
gl_stop and fill_gl_memory vfuncs, along with properly advertizing the GL APIs
it supports through the supported_gl_api GstGLBaseSrc class attribute.
Commit c71dd72b "gl/wayland: fix glib mainloop integration" was overeager
in removing the poll result test from the check function. This caused
dispatch to be called even if no new events are available on the
Wayland connection, which in turn would wake up the glib mainloop,
causing effectively a tight loop without ever blocking on the poll.
Fixes#603
We use this property in gst_gl_display_egl_from_gl_display, to set
foreign_display for the new GstGLDisplayEGL instance. This fixes a
problem where gst_gl_display_egl_finalize calls EglTerminate on a
pre-existing EGL connection.
It seems that eglCreatePlatformWindowSurfaceEXT is failing (with
EGL_BAD_ALLOC) because it thinks an EGL surface has already been created
for the wl_egl_window. The reason is that the "driver_private" field of
the wl_egl_window is getting clobbered by the function
wl_proxy_set_queue().
Since a wl_egl_window is not a wl_proxy, it shouldn't be passed to
wl_proxy_set_queue(). It just wraps a wl_surface (which is a wl_proxy).
And it looks like the queue for that surface is getting set earlier on
in the function anyway.
See https://gitlab.freedesktop.org/gstreamer/gst-plugins-base/issues/621#note_184582
