Here is some additional input that may help teams.
We had:
Code:
#define UINT32 unsigned int
in one of our files. This is actually slightly different than the type defined in WPILib (even though it eventually resolves into an unsigned int). The compiler/linker could successfully resolve that two-level symbol, but when placed on the cRIO, the variable types were slightly different and so the symbols did not match up.
This is not an obvious problem, with no obvious error message given. So other teams may have encountered this in some form also.
And Windriver/Eclipse guys reading this... why do you filter out -pedantic from the g++ Makefile settings in Eclipse IDE? It is very useful, and may have even caught this error.
Thanks to everyone for their help and input. Figuring out problems like this will be very useful for a lessons learned on the new cRIO controller.