Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project! See Demo

Close

#100 "Too many eyesplits" triggered with displacementbound >= 4.

open-accepted
nobody
None
3
2008-10-19
2008-10-16
Adam Buchbinder
No

Running the attached RIB under Pixie SVN (r1169) throws a very long series of "Too many eyesplits" errors.

Is this an architectural problem with the way Pixie is structured, or it bailing out of the eyesplit routine too early? I wouldn't have filed a bug on this, except that other renderers work properly with it--BMRT renders it, as does 3Delight; Aqsis gives a similar error "WARNING: Max eyesplits for object "unnamed" exceeded". Moving the camera back and/or decreasing the displacement bounds allows the render to proceed.

The original files are from http://www.edit.ne.jp/~katsu/rms_tree.htm; I fixed the line endings so they'd be interpretable and hacked it down to test-case size. It originally included some astonishing abuse of the shading language, but that's been edited out as well in the included sample.

Discussion

  • RIB triggering the error.

     
    • priority: 5 --> 3
     
  • Okan Arikan
    Okan Arikan
    2008-10-19

    This is an architectural issue related to scan line rendering. If you change the hider to "raytrace", it will render without the eye split warnings. Overall, Pixie may be a little overzealous in the warning messages here. I will implement a cleaner detection of problem cases.

    Looking at the RIB, the scene seems quite simple and I do not see a cause for the eye split warnings.

     
  • Okan Arikan
    Okan Arikan
    2008-10-19

    • status: open --> open-accepted