On May 2, Keith Packard wrote: > Around 11 o'clock on May 2, Carl Worth wrote: > > > This is what had been thinking to do originally. Do you see any > > problem with this approach? > > Yes. Which coordinates get snapped determines the result of the > computation; this essentially codifies one implementation as the standard > rather than building a spec which can be matched with multiple different > implementations. In particular, I doubt very much that we'd ever see > hardware match this exactly.
What if the implementation corrected for the snapping by adjusting the area calculation by adding/subtracting thin rectangles whose height was determined by the error left from the bresenham snapping? The inaccuracy from that would come from the fact that one end of the error rectangle should actually be clipped by the angle of the line. That should give a maximum inaccuracy of 1/2 of one sub-pixel per error rectangle. And, each pixel would need two error rectangles, so we have a maximum error of one sub-pixel. Good enough? -Carl _______________________________________________ Render mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/render
