late expose events may corrupt display

[dylex]

It is possible that an expose event and a draw call get reordered so that the expose ends up drawing on top of the already drawn display. One blunt solution would be to always clear on expose (research this). Alternatively, we could force processing of exposes before any draw, but this doesn't seem complete or efficient. It's also quite possible this doesn't hurt anything anyway, since it should draw the same thing. The exception being the cursor, and I suspect this may be causing a cursor-wrong-on-attach bug, but I'm not sure.

[dylex]

That was not the cursor-attach bug, which is now fixed in 96711a5f9bea685774c119c5b2f5c99cacb1f56e. I have not yet seen any actual evidence for this bug, so I suspect it may not actually matter (at least without xft or something with antialiasing).