Le Tue, 18 Jun 2002 11:48:31 +0200 (CEST) "Jun P.FURUSE"
<Jun.Furuse@inria.fr> a écrit :


> I verified that this code prints out "BUG" when X display has less
> than 24bits color depth. In such environment, the colour mapping from
> Graphics to X is not one-to-one. This means that point_color may not
> always return the exact colour which you drew on the screen. 
> For example, in an X display of 15bpp, the Graphics colour 0xffffff
> and 0xf8f8f8 are translated to the same X white colour. 
> This is also true for the default white background color of Graphics. 
> point_color returns 0xf8f8f8, not 0xffffff, this is why the above
> code prints out "BUG".

OK. Now I have this simple problem : I can use Direct Rendering from X
with my Voodoo card in 15 bits, and correct color translation in 24 bits.
is that right ?
 
> In Camllight, the translation between Caml colour and X colour
> performs communication between programs and the X server. Since it is
> highly expensive operation, the colour translation of O'Caml Graphics
> is rewritten, so that it would never ask the X server about
> colours. It improves the speed of colour query impressively. 
> Unfortunately, the both colour translation code do not return always
> the same result. The different behavior of the Nicolas' O'Caml and
> Camllight programs can be explained by this difference of colour
> translation.

Who said computer science is not an exact one ?
 
> I think this does not mean the new colour translation of O'Caml is
> buggy. Even in Camllight, the same things can happen, unless the colour
> translation is one-to-one. (It happens less, but not never.)
> 
> The important things (which should be noted in graphics.mli) are that
> 
>   * you may get different colour index from one you draw, 
> 
>   * the colour indices predefined in Graphics for the basic colours 
>     such as white, blue, red, etc can be just canonical candidate
>     indices for them.
> 
> You should not compare these colour indices against the colours
> got from the screen. For example, in Nicolas' program, instead of
> comparing Graphics.white, we should use the colour index obtained
> from the screen  at the begining of the program:
> 
>       let screen_white = point_color 0 0 
>       (* this can be different from Graphics.white *)

Great. Thanks for the tip. Euh, and what if I have to test several colors
? Oh, yes, here it is :

let screen_from_color c =
  set_color x;
  plot 0 0;
  point_color 0 0;;
 
Right ? I'll try it now.

\bye, and thanks again.

-- 

                   Nicolas FRANCOIS
            http://nicolas.francois.free.fr
 A TRUE Klingon programmer does NOT comment his code
-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners