[Xcb] XCBImage

[Jamey Sharp]

On Sun, Mar 06, 2005 at 12:05:37PM -0800, Jamey Sharp wrote:
> I have some ideas for simplifying the Xlib-derived code, which should
> make it run faster, use less memory, and be easier to validate. I won't
> get into those just yet, though; we'd better make sure the current stuff
> works first, I think.

I changed my mind. I've just committed a new version of xcb_image.c with
a re-write of get_pixel/put_pixel bits. A slight relaxation of
invariants let me throw away a ton of code that was trying to maintain
those invariants. I also removed all the specialized, "optimized"
versions so we can focus on correctness of a single implementation of
each operation on each image format.

I think I might actually understand XYPixmap and ZPixmap formats now,
and I hope the put_pixel and get_pixel functions are clearer. As an
added bonus this cuts compiled code size roughly in half.

A thorough test app would be really nice. Maybe it could create an
XCBImage and a PIXMAP for every depth on the default screen, do PutImage
and GetImage requests in both XYPixmap and ZPixmap formats, and walk
across the resulting images with get_pixel comparing values. I suppose
the same app could do extra tests on whichever depth/format is supported
by MIT-SHM.

--Jamey