[Radiance-general] rpict can recover, can rpiece?

[Greg Ward]

Hi Mark,

The "bad ray direction" is a serious error, caused either by a 
miscompiled binary or a corrupted ambient file.  Assuming it's the 
latter, you should try deleting (or renaming) your ambient file before 
attempting recovery.  A corrupted ambient file is usually the result of 
a faulty NFS lock manager.  See previous posts on this topic.

-Greg

> From: Mark Stock <mstock at umich.edu>
> Date: May 27, 2004 8:25:40 AM PDT
>
> I am rendering a medium piece (8k x 8k) using rpiece on a
> cluster of Opteron processors, and I have used "rpiece -F"
> to create the sync file and start 4 processes. I even used
> "rpiece -R" to restart the job once. (Thanks to Giulio
> Antonutto for helping me get started.)
>
> Anyways, there are two parts of the 64 subdivisions of the
> image that are giving me trouble. I invoke them piece-at-a-
> time:
>
> % more argsN
> -X 8 -Y 8 -t 600 @vp11 @opts -x 8000 -y 8000 -o img13.pic scene.oct
>
> % echo 2 0 | rpiece @argsN
> rpict: warning - no light sources found
> FRAME 1: -vta -vp 1.3 -1.8 1.1 -vd -0.524211 0.708567 -0.472373
> -vu 0 0 1 -vh 5.45455 -vv 5.45455 -vo 0 -va 0 -vs -1.5 -vl -3.5
> rpict: 0 rays, 0.00% after 0.000u 0.000s 0.000r hours on str2
> rpict: consistency - bad ray direction in inithemi
> rpiece: read error from rpict
> Command exited with non-zero status 1
>
> I understand the "no light sources" error: I illuminate with
> the sky dome only. It's the "bad ray direction" that is causing
> me problems.
>
> Now, I had previously seen this error when I rendered a smaller
> version with rpict. I simply re-ran rpict with the "-ro" recover-
> overwrite option. But it seems that the "rpiece -R" recover
> option restarts any incomplete subsections instead of continuing
> where rpict left off. So, every time I try to patch the missing
> pieces, I get the same error at (presumably) the same place.
>
> So, can rpiece recover in a pixel-wise sense like rpict, or,
> how can I find out what command sequence (including rpict and
> not rpiece) can render those bad patches and put them in the
> final image.
>
> Or, is there any rpiece option that will ignore the "bad ray
> direction" error, or mitigate its effects?
>
> Thank you.
>
> Mark