[MACTCL] brainstorming TkMacOSXUpdateClipRgn()

[james t. <ti...@ma...>]

ok,

...it's an old problem, but unsolved, and even if we move to CGPaths 
instead of QD regions, it is key to some of our performance woes:  
everytime something is drawn in a window, it is invalidated and then 
the window and all of it's subwindow's are recursively re-examined to 
update their clip region...this is very slow if much is embedded in the 
window, and I believe it's been mentioned before that it was done 
because it was wholly correct (perhaps too so)...

...atm, TkMacOSXUpdateClipRgn() is only called in two places (outside 
of itself):  XMapWindow(), and TkMacOSXSetUpClippingRgn()...this last 
one is called all over the place where there is drawing...

...so, thinking out loud, what are some alternatives?  The only thing 
I've come up with is that instead of rechecking every single embedded 
subwindow/object top-down, we should have some mechanism for weeding 
out those child windows that don't intersect with the drawing:  does 
this sound plausible?

...does anyone have other ideas, or obvious reasons why something like 
this wouldn't work?

jamie