I think the problem here is more likely to be this bit: On Thu, 29 Jun 2023 at 16:21:36 +0200, Paul Gevers wrote: > 228s libEGL warning: failed to get driver name for fd 0 > 228s 228s libEGL warning: MESA-LOADER: failed to retrieve device information > 228s 228s libEGL warning: failed to get driver name for fd 0 > 228s 230s [02:18:06.642] caught signal 15
and "panel surface gone" seems more likely to be a symptom of that than the cause? GTK's test suite is relying on being able to do EGL under Wayland on a headless machine, at least via llvmpipe. smcv