I wonder if there are some good practices for testing multiple configurations in Hudson. We compile tests in debug and release mode and also toggle MPI on and off.
The make system takes a series of parameters (DEBUG=FALSE, OPT=TRUE, MPI=TRUE) and does the right thing, producing test executables that are well labeled. I could probably create a monolithic control script that does everything, but I wonder if I am missing some facility for iterating through a list of parameters. Even better would be if I could clean the project in between.
Anyone have an opinion? Is this best done with separate projects? With parameters? Just by writing a very complete script?