Around 14 o'clock on Mar 29, Vladimir Dergachev wrote:
> > The Gatos code was a bit broken, working
> > for only depth 16 and depth 24 TrueColor visuals, but that was easy enough
> > to fix.
>
> Keith, could you tell me what was broken in GATOS code ?
RADEONSetColorKey needs to convert a pixel value into a color, and then
detect the valid bits in each component. The GATOS code had two cases,
a 565 TrueColor case and an 888 TrueColor case. Pseudo color, 555
TrueColor or any other visual class wouldn't get the right values.
Take a look at the patch I posted yesterday at
http://keithp.com/~keithp/download/radeon_xv.diff
to see what that version of RADEONSetColorKey does. I'm not completely
happy with that code either; it assumes the LUT outputs will precisely
match the X colormap values for non TrueColor visuals; I doubt that is
correct in all cases. I suspect that the code should check for 6-bit DACs
for Pseudo/Static visuals and also use the weights for Direct color.
More testing is clearly needed...
Keith Packard XFree86 Core Team Compaq Cambridge Research Lab
_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert
