Comment 13 for bug 708388

Being a pain, I'd like to see some profiling to further identify the cause.

A lot of the work I'm hoping to do relies on invalidating objects "just working", and redraw being fast without any trickery. By invalidating a big rectangle first of all, I assume GTK will then not add separate small rectangles when they lie within the already invalidated region. Our expose event will be called with the big rectangle and we draw once.

In the current case - where oes the time go? Is it spent in GTK adding the regions to its data-structures? Or is it that GTK hands us an expose event with lots of small rectangles - and we then spend ages drawing?

Could it be that we're hitting performance issues in cairo when rendering with a complex clip region?

If the rendering is to blame, we could choose to redraw the entire rect bounding the damaged regions - without having to fake anything by explicitly damaging the whole region first.