Re: [Mesa3d-dev] Inconsistent scan conversion in XOR mode

[Howard <ho...@me...>]

----- Original Message -----
From: "Brian Paul" <br...@pr...>
To: "Howard" <ho...@me...>
> >
> > Has anyone else observed a problem with the scan conversion of vertical
and
> > horizontal lines while in XOR mode. We have consistently observed that
there
> > are:
> > a) Irregularities in vertical and horizontal line, i.e. 4 - 10 pixel
spans
> > that are offset by a pixel.
> > b) Non repeating patterns, i.e. the same vertical or horizontal line
> > coordinates plot these irregularities differently. This is especially
> > visible/troublesome in the XOR case, because the 2nd, (erase) draw does
not
> > cover the same pixels, and consequently leaves scars. This only seems to
> > happen in the vertical/horizontal cases and does not happen when in a
> > normal, (non XOR) mode.
>
> I've seen a problem in which a clipped, vertical, wide line will "wiggle"
> left/right by a pixel.  When I looked into it it was a case where one
> endpoint was at X=5 and the other at x=4.999 (for example).
>
> Are your endpoints being clipped?
>
No, there is no clipping. There must a round-off of some kind. The data
points are the same, but the view matrix could be slightly different. The
funny thing is, the start and end points will be level, but the path, as you
say, "wiggles", when in XOR mode. The problem is not present with the same
code on Irix. We will try to distill this down to a small example if no one
else is experiencing it.

Thanks,
--
Howard Luby     Fax:248 540-3138 Tel:248 540-2251
ho...@me... http://www.mediascape.com