Trapping runtime errors

[ereisch]

Is there a way to have the forrtl error handler call a user-generated program after it prints out its stack trace? We have a multi-process environment and when one process crashes due to a forrtl error, we want to be able to signal the other processes that the one has crashed, so they can deal with it appropriately, rather than blocking indefinitely on a process that no longer exists. Even if it did something as simple as raise a signal, we can install a signal handler to trap that and react accordingly.

Thanks

[Steven_L_Intel1]

Look at the documentation for the SIGNALQQ library routine.

[ereisch]

Thanks, I saw that, but when forrtl encounters a "generic" severe error, is there a signal raised? If so, which one? An example:

forrtl: File exists
forrtl: severe (10): cannot overwrite existing file, unit 6, file /path/to/somefile.txt


Is there documentation somewhere (similar to http://software.intel.com/sites/products/documentation/hpc/compilerpro/en-us/fortran/lin/compiler_f/bldaps_for/common/bldaps_rterrs.htm) that lists which signals are sent for which errors? In the above link, it says that "to prevent program termination...change the default action of a signal", but isn't clear on which signal we should be trapping. FPEs are understandable, as well as SEGVs, but what about arrays out of bounds?

Thanks again

[Steven_L_Intel1]

Unfortunately, I/O errors are not signalled and can't be trapped. I have previously filed a feature request asking for that ability - I will add yours to the list.