Allow turning off trap_fpe in debug mode

[permcody]

closes #3595

@shanestafford this should enable you to use FP_INFINITY in your tests, just make sure that you using the cli_args option if you create a test that will run in debug mode to pass the --no-trap-fpe flag.

[moosebuild]

Results of testing aa0dacf using moose_PR_pre_test recipe:

Passed on: linux

View the results here: https://www.moosebuild.com/view_job/3975

[moosebuild]

Results of testing aa0dacf using moose_PR_test recipe:

Passed on: linux

View the results here: https://www.moosebuild.com/view_job/3976

[jwpeterson]

I still don't think you are supposed to use variables with two leading underscores...

The rules (which did not change in C++11):

Reserved in any scope, including for use as implementation macros:
* identifiers beginning with an underscore and an uppercase letter
* identifiers containing adjacent underscores (or "double underscore")
Reserved in the global namespaces:
* identifiers beginning with an underscore
Also, everything in the std namespace is reserved. (You are allowed to add template specializations, though.)

reference

[moosebuild]

Results of testing 3f53fcb using moose_PR_pre_test recipe:

Passed on: linux

View the results here: https://www.moosebuild.com/view_job/3996

[jwpeterson]

@shanestafford can you try this branch, make sure it does what you want?

[permcody]

@shanestafford in fact, I wouldn't mind having a simple test added to moose/test. We had a nan_test in misc/check_error but it has been skipped since Petsc 3.4 because PETSc catches nans now. The FP_INFINITY capability should probably be tested at the MOOSE level to ensure that it continues to function. Feel free to contribute the computation that you want tested.

[moosebuild]

Results of testing 3f53fcb using moose_PR_test recipe:

Passed on: linux

View the results here: https://www.moosebuild.com/view_job/3997

[shanestafford]

I'll add it to the hydride test and report back.

[shanestafford]

Status update: the branch does fix the test that failed (linux-gcc on a test that is not in MOOSE).

However, testing this change in MOOSE is proving to be difficult since most of the exp(-A/T) functions are in material models outside of MOOSE. Apparently, the parsed function does not respect the same fpe flags, so it is not possible to do something trivial like a FunctionAux with function = 'exp(-1/t)' to test this. There are some exp functions inside of solid_mechanics that I will try next.

[permcody]

Feel free to muck with one of the test material objects in moose/test/src/materials, that's what they're there for. Just add another flag to trigger the desired computation on demand. We already have other tests in the suite that do this.

[permcody]

Closed in #3649