[Bug 785518] rpi, gl: eglCreateImageKHR failing

[GStreamer (GNOME Bugzilla)]

https://bugzilla.gnome.org/show_bug.cgi?id=785518

--- Comment #5 from Julien Isorce <julien.isorce at gmail.com> ---
> (In reply to Julien Isorce from comment #3)
> > Maybe not enough gpu mem by default, can you try to set gpu_mem=512 in
> > /boot/config.txt ?
> > 
> > In order to use less gpu memory (FBO) there is a GstGL bug tracker here
> > https://bugzilla.gnome.org/show_bug.cgi?id=730927
> 
> Indeed, that did the trick thanks!

Great! Could you check if the assert "CRITICAL **: gst_egl_image_get_image:"
comes from the call of gst_gl_memory_egl_get_image from gstomxvideodec.c ? If
yes then there is a bug in GstGL because it means that
gstomxvideodec.c::gst_buffer_pool_acquire_buffer succeeded. So I think GstGL's
gst_gl_buffer_pool_alloc should not return GST_FLOW_OK in this case.

Also I wonder how we could improve the out of mem detection.
There is the message "eglCreateImageKHR:  failed to create image for buffer 0x1
target 12465 error 0x300c", but it is not from GstGL right ?
0x300c -> EGL_BAD_PARAMETER

Ideally it should have return EGL_BAD_ALLOC
"If insufficient memory is available to complete the specified operation, the
error EGL_BAD_ALLOC is generated."
So that it points to the action which is to increase mem.

Also eglCreateImage should only increment a reference counter so it could be
that gstglmemory.c::_gl_tex_create could have catch GL_OUT_OF_MEMORY in the
first place.

> 
> Should I close this as a dup of bug #730927 then?

I am not sure because that other bug mentions that this is for the YUV cases.
Can you count how many times gl->GenFramebuffers is called ? If 0 then not a
duplicate.
(also note that the hw decoder also uses some gpu mem)

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.