On 1/25/19 6:13 AM, gnagflow wrote:
> Hi,
> although hidden behind the red plane, parts of the blue plane can be seen
> from the following location.
> 
> -------------- cut here ----------------------------
> #version 3.7;
> 
> global_settings {
>    assumed_gamma 1.0
> }
> 
> camera {
>    perspective
>    location     -1 * <0, 1, 1/50000>
>    direction     1 *  y
>    look_at            y
> }
> 
> light_source
> {
>      <0, -1, -1>
>      color red 1.0  green 1.0  blue 1.0
> }
> 
> plane { z, 0
>    pigment{
>      color rgb <1, 0, 0>
>    }
> }
> 
> plane { z, 1
>    pigment{
>      color rgb <0, 0, 1>
>    }
>    finish {
>      ambient  1
>      emission 0
>    }
> }
> ----------------------------------------------------
> 
> The effect was discovered on Linux kernel 4.20.2 with x86_64 GNU/Linux
> architecture using POV-Ray 3.7.0.8.unofficial version but can pretty much be
> reproduced on any system as far as I can tell. Accept my apologies if the
> issue is already known, in which case I would be grateful for any further
> hints regarding a remedy.
> 
> Regards
> Wolfgang
> 
> 
> 
Looks to be related, by magnitude, mostly to the work going on for this 
pull request:

https://github.com/POV-Ray/povray/pull/358

and two or three similar open github issues - most/all? mentioned in the 
358 discussion thread as I recall.

In your case the first ray surface intersection is being ignored if it 
is closer than 1e-4. FYI. I'm adding your scene to my solver/tolerance 
testing scene collection.

The current - and near term - work around in your scene is to keep the 
camera more than 1e-4 off/outside the red plane's surface.

A better rule of thumb though is >1e-3 outside shapes / away from 
surfaces for lights, cameras and surface relationships - in the world 
scene space. In some scenes with the orthographic camera & user_defined 
cameras that might not be possible depending upon the scene. It's not 
enough for some shapes and scene set ups, but still a decent rule of 
thumb value.

My current, local, solver branch going after a set of related fixes 
renders only the red plane for what it's worth. That branch is an effort 
aimed at fixing/improving a largish set of accuracy issues to which this 
one (pull 358 et al) is somewhat related.

Bill P.

Post a reply to this message