Bugzilla

(In reply to info@michaelzeilfelder.de from comment #7)
> Created attachment 9203062 [details]
> about:support values
> 
> I don't think I enabled anything manually - at least not recently. And not using Wayland.
> 
> But it may be related to es2 not working on Debian anymore. Thought that isn't something new, but has been broken for a long time by now. Don't think it ever worked with Debian 10, I think it was even broken in Debian 9 (I tried to report that to them, but unfortunately Debian bugtracker is not as nice as your bugzilla...  so I failed sending them the bugreport several times and then gave up). For example starting es2gears on Debian I get: libEGL warning: DRI2: failed to authenticate
> 
> But WebGL still worked in Firefox for some reason. And still works even now in Chromium.

WebGL should work fine on top of desktop GL. But I think I now see what's wrong: you're probably running into a fallback method that we use to discover the right GPU - even going the EGL path when we originally started with GLX. Now it this case it fails in the EGL part, making the whole test fail. You you test if you still see that in beta and nightly? That would be really good to know.

(In reply to info@michaelzeilfelder.de from comment #7)
> Created attachment 9203062 [details]
> about:support values
> 
> I don't think I enabled anything manually - at least not recently. And not using Wayland.
> 
> But it may be related to es2 not working on Debian anymore. Thought that isn't something new, but has been broken for a long time by now. Don't think it ever worked with Debian 10, I think it was even broken in Debian 9 (I tried to report that to them, but unfortunately Debian bugtracker is not as nice as your bugzilla...  so I failed sending them the bugreport several times and then gave up). For example starting es2gears on Debian I get: libEGL warning: DRI2: failed to authenticate
> 
> But WebGL still worked in Firefox for some reason. And still works even now in Chromium.

WebGL should work fine on top of desktop GL. But I think I now see what's wrong: you're probably running into a fallback method that we use to discover the right GPU - even going the EGL path when we originally started with GLX. Now in this case it fails in the EGL part, making the whole test fail. Could you test if you still see that in beta and nightly? That would be really good to know.