|
 |
Hello,
I think that I have come across a bug in POV-Ray. Try to render the
following scene:
background { color rgb 1 }
camera {
sky z
location <1757,792,798>+<0,800,0>
look_at <1757,792,798>
}
light_source { <-50,800,0>+<1757,792,798> color rgb 1.5 }
bicubic_patch {
type 1 flatness 0.0 u_steps 4 v_steps 4
<1277.46,645.096,730.238><1273.35,640.857,678.616><1271.38,636.003,630.537><1271.54,630.534,585.999>
<1275.19,678.859,731.733><1315.94,664.298,681.323><1317.52,659.516,632.745><1279.95,664.515,585.999>
<1323.73,702.723,730.807><1358.35,687.334,681.949><1355.24,682.46,633.679><1314.41,688.101,585.999>
<1423.09,716.687,727.459><1400.58,709.966,680.493><1384.54,704.834,633.34><1374.95,701.292,586>
pigment{color rgb 1}
}
If you replace "type 1" by "type 0". The patch will look quite different.
In my opinion the surface normal is not correct. Can some one confirm that
this is a bug and check whether it is corrected in version 3.1b? (Since
I use Unix I can not do it.) I will then email a bug report to Chris.
Thomas
--
Tho### [at] uni-konstanz de
http://www.fmi.uni-konstanz.de/~willhalm/
Tschieses lavs ju
Post a reply to this message
|
|
|
|
On 12 Aug 1998 14:49:14 +0200, Thomas Willhalm
<Tho### [at] uni-konstanzde> wrote:
>
>
>If you replace "type 1" by "type 0". The patch will look quite different.
>In my opinion the surface normal is not correct. Can some one confirm that
>this is a bug and check whether it is corrected in version 3.1b? (Since
>I use Unix I can not do it.) I will then email a bug report to Chris.
I can confirm that the bug still exists in the win32 build of beta 5. If you
have the 3.0x sources, here's the fix:
In the file bezier.c in the function bezier_subdivider, lines 1526, 1538,
and 1545-1546 look like this:
1526: vt = (v1 - v0) / 2.0;
...
1538: ut = (u1 - u0) / 2.0;
...
1545: ut = (u1 - u0) / 2.0;
1546: vt = (v1 - v0) / 2.0;
These lines don't match the similar lines in bezier_tree_builder, where all
of the minuses are replaced with plusses. Indeed, replacing the minus signs
in the above four lines with plus signs fixes the problem.
I'll let you go ahead and send the bug report to Chris. If you'd like to
quote this post, please feel free.
Post a reply to this message
|
|
|
|
par### [at] my-dejanewscom (Ron Parker) writes:
> On 12 Aug 1998 14:49:14 +0200, Thomas Willhalm
> <Tho### [at] uni-konstanzde> wrote:
> >
> >
> >If you replace "type 1" by "type 0". The patch will look quite different.
> >In my opinion the surface normal is not correct. Can some one confirm that
> >this is a bug and check whether it is corrected in version 3.1b? (Since
> >I use Unix I can not do it.) I will then email a bug report to Chris.
>
> I can confirm that the bug still exists in the win32 build of beta 5. If you
> have the 3.0x sources, here's the fix:
>
> In the file bezier.c in the function bezier_subdivider, lines 1526, 1538,
> and 1545-1546 look like this:
>
> 1526: vt = (v1 - v0) / 2.0;
> ...
> 1538: ut = (u1 - u0) / 2.0;
> ...
> 1545: ut = (u1 - u0) / 2.0;
> 1546: vt = (v1 - v0) / 2.0;
>
> These lines don't match the similar lines in bezier_tree_builder, where all
> of the minuses are replaced with plusses. Indeed, replacing the minus signs
> in the above four lines with plus signs fixes the problem.
Thank you very much! This was a fast and very helpful response!
> I'll let you go ahead and send the bug report to Chris. If you'd like to
> quote this post, please feel free.
This is what I'll do.
Thanks again
Thomas
--
Tho### [at] uni-konstanzde
http://www.fmi.uni-konstanz.de/~willhalm/
Tschieses lavs ju
Post a reply to this message
|
|
