nanogui: Thread: Microwindows - Back to Basics

Does any one know of a better way to find out what part of a partially
covered window to update
other than calculating all of the clip rectangles? This "Basic" question
has bugged me since the
late 80's and I still have not found an "easy'" way to do it. Back then
I could fool most of my
customers into thinking that their new hardware ran on a "Real"
windowing system by just following
the simple rule (read scam) that only the infocus window can be updated
and only the top window
can be in focus. A lot of that hardware is still around and most of the
end users still have not caught
on.  Drawing routines get very fast if you can ignore all those clip
points. In 90 or 91 I bought one
of those expensive (at that time) 3rd party graphics tool kits that
supported lot's of hardware and
drawing to virtual screen buffers. I used this new library as the base
of my new windowing system
for intel hardware but I still had the problem of  "what to update" when
it came time to repaint the
screen. My memory hog of a solution (Remember even 4 meg was a lot back
then) was to draw all
windows to ram virtual buffers and at repaint time check for overlap on
upper windows and then
blast all need windows to the screen. Since the screen was being double
buffered the extra repaints
from overlaps was hidden. Since drawing to ram in 256 color mode was
very fast the only real
waste - other than all the ram needed - was in the repainting of
overlapping windows. The benefits
of the system from my point of view  - it was simple - the logic of
figuring out what to send
to the screen was a no brainer. It also didn't suffer from the oh-oh I
just ran out of clip rectangles
problem.  Unfortunately the real world came into play again and the
windowing system had to
run on a V25 cpu with only 256k of static ram and an EGA card. Back to
counting rectangles
again. Its now 10 years later and we are still counting rectangles. Is
there a better way?

Its now 10 years later and we are still counting rectangles. Is
: there a better way?


Not that I know of.  I spent a good month making sure that all the rectangle 
management is performed properly in NanoX and Microwindows, and we're
still not done.  BTW, this method uses a heck of a lot less ram than
the method you were using in '91.  All this stuff computes rectangles
for every graphics context change, and it still runs in ~ 40k.

Greg