Re: Testing for runtime errors

[Janis Johnson <janis187 at us dot ibm dot com>]

On Fri, May 05, 2006 at 05:02:10PM -0700, Jerry DeLisle wrote:
> Janis Johnson wrote:
> >On Fri, Apr 28, 2006 at 10:37:33AM -0700, Janis Johnson wrote:
> >
> >>On Thu, Apr 27, 2006 at 09:38:35PM -0700, Jerry DeLisle wrote:
> >>
> >>>(Andrew Pinski suggested I cc Janis on this)
> >>>
> >>>Do you have any suggestions on how to test for run time error messages 
> >>>in the gfortran testsuite?  When there is a runtime error, the test 
> >>>program exits with "1" so the dejagnu magic interprets it as a failure.
> >>
> >>I'll take a look next week.  Ping me if you don't hear anything.
> >
> >
> >Try this.  It provides a new test directive, dg-nonzero-exit, which
> >takes an option target specifier and option lists.  I think it does what
> >you need for Fortran library testing, with broader support because it
> >will probably also be useful for other functionality that requires tests
> >whose executables are expected to return nonzero results.  I need to do
> >further testing to make sure I didn't break dg-skip-if, and to test
> >"make dvi" and "make info".
> >
> 
> The concept looks great!  I have applied the patch and I am bootstrapping 
> now to test.  Will report back when done.

I'm more interested in having you try using it in Fortran runtime tests
to see if anything else is needed; I plan to test it early next week.

Janis