Erik Kurzinger

Could you please try with `WAYLAND_DEBUG=1`? That should give more information about the source of the protocol error.

@hasheddan I think you truncated the output right after the `zwp_linux_buffer_params_v1_add` request which would be helpful to see. Mainly the last two parameters which are the DRM format modifier.

Thanks! It looks like the log for `zwp_linux_buffer_params_v1_add` got scrambled up with another event, but it's on line `[3039631.100]`. So the format of the failing buffer if DRM_FORMAT_ARGB8888 and the...

Looking at the `choose_primary_gpu_unchecked` function in the mutter code-base, it seems that it will use the boot VGA device by default, or an arbitrary device if none of them have...

Could we perhaps test that theory by simply swapping the two cards?

The only other thing I can think of would be to apply the tag to the render node (/dev/dri/renderDXXX) instead of or in addition to the primary node (/dev/dri/card1). If...

It looks like the code added by this commit https://mail.gnome.org/archives/commits-list/2016-May/msg04276.html doesn't interact well with the buffer release thread added in 1.1.8. Qt had a somewhat similar issue, which was eventually...

> There seems to be a missing wl_display_flush() before polling in wlEglHandleImageStreamEvents() Did you mean in buffer_release_thread? If so, yeah, I noticed that too, but unfortunately adding it doesn't seem...

After spending some more time trying to make sense of this, I have a slightly better idea what's going on, although still not entirely clear. GTK's frame scheduling / event...

To clarify, the mentioned hack was in GTK's code, not egl-wayland, and I have a feeling it would break other things. It was more of an experiment than a real...