Get Sub Matrix returning NAN

[jab0707]

I ran into a weird bug using get submatrix. In the module below if I set get matrix slice to 0 (which will generate volume indices [0,1,2] the output matrix is full of nan/really high values. If I change to any other slice (any other set of three columns) then I get the expected output. Network attached for reference. Does not require external data.

version v5.0-beta.Z
WeirdBug.srn5.zip

[dcwhite]

I replicated this, it's reading from uninitialized memory so you'll get NaNs and garbage out. Very odd that it's a simple function of SelectSubMatrix that does this, as there are plenty of tests written for that module (not by me though :). It will be an easy fix.

Also, the "WeirdBug.srn5" reminds me of the "one weird trick" ads around the web: https://en.wikipedia.org/wiki/One_weird_trick_advertisements

[dcwhite]

This line is the culprit:
https://github.com/SCIInstitute/SCIRun/blob/master/src/Core/Algorithms/Math/SelectSubMatrix.cc#L349
It's only selecting columns when the index is > 0, so the 0th column in the output is random garbage. This sort of copy-paste bug is the result of the heavily-nested code of the algorithm, trying to handle too many cases in a simplistic manner.