Code review comment for lp:~alan-griffiths/mir/fix-1695221

> Without testing this, the fix sounds reasonable.
>
> Using GL in a Mir callback is expected to fail. Even if you remember
> context.make_current() and even if it works, I recall you could still hit
> deadlocks making Mir/GL calls from in that event callback thread.

That's exactly what was happening here.

> P.S. mir_demo_client_target is intended to replace and supersede
> mir_demo_client_eglsquare. See also bug 1433925 and bug 1651695, which we
> probably don't care to fix. Those bugs don't exist in mir_demo_client_target.
> As such I wouldn't spend too much time trying to fix eglsquare - it's just too
> broken and not necessary.

I spent the time because I wanted to understand the problem. Fixing was a quick confirmation of the diagnosis.

(Happy to see eglsquare deleted.)