new g77 testuite cases. help needed.

new g77 testuite cases. help needed.

[Robert Lipe]

I've been picking through archives of the list snaring problem reports.
I remember very little about Fortran.   I have boiled down several test
cases, but would like to work with a g77-compatible person to get them
right.   I'm not sure that I captured the spirit of all the tests.

In the g77 testing framework, how can I tell it "this test needs to be
compiled with '-fPIC' or with '-g' or other options"?

I'll start committing these new cases soon.   

-- 
Robert Lipe       http://www.dgii.com/people/robertl       robertl@dgii.com

Re: new g77 testuite cases. help needed.

[Jeffrey A Law]

  In message < 19980311000222.21501@dgii.com >you write:
  > I've been picking through archives of the list snaring problem reports.
  > I remember very little about Fortran.   I have boiled down several test
  > cases, but would like to work with a g77-compatible person to get them
  > right.   I'm not sure that I captured the spirit of all the tests.
  > 
  > In the g77 testing framework, how can I tell it "this test needs to be
  > compiled with '-fPIC' or with '-g' or other options"?
  > 
  > I'll start committing these new cases soon.   
The existing g77 testsuite is 100% based on the c-torture framework;
there's not "gcc.dg" framework for the g77 testsuite.

Therefore, there's no good way to specify additional options for
the testsuite.  c-torture and f-torture are really designed for
tests which we don't need to do anything special for.

Now, having said that, there's a backdoor you can use to override
much of the c-torture framework -- the ".x" files are run if they
exist instead of using the standard c-torture testing framework.

So in theory you could use them to specify additional options or to
perform some other hackery.

jeff

Re: new g77 testuite cases. help needed.

[Dave Love]

I can't drive dejagnu, but would offer advice on test cases, time and
knowledge permitting, preferably in circumstances where Craig/Toon had
the opportunity to catch errors.  It might be best to do this via the
fortran bug address -- thoughts from them?