nist testsuite question.

[Jerry DeLisle]

Uttam Pawar wrote:
> Hi All,
> 
> I was looking through the NIST fortran testsuite which as we know is
> written for fortran77. 
> I'm having problem in compiling and running them with gfortran with
> -std=f95 option. The construct in a test specially arithmatic IF
> statement (which checks the result of computation) with multiple go to
> is causing the tests not to compile with -std=f95 (obviously).
> 
> Is there any plan on converting them or writing new tests to make them
> compatible with f95 standard?
> 
> Can we really convert them and add to the gfortran test suite?
> 
There is no real point in converting NIST to f95.  One of the many reasons we 
test with NIST is to make sure we can compile F77 code with gfortran since 
gfortran is the replacement for g77.  I will be putting together automatic 
regression testing of NIST sometime in the not too distant future.

I don't think we want to add it to the regular testsuite because it adds more 
time to testsuite execution which slows down development cycles. (IMHO)  I think 
that time would be better spent focusing on newer f95 features.


> In NIST testsuite, there are no tests at least for sinh() and cosh()
> functions with negative arguments, should we add them as new to gfotran
> tests?
> 
This is debatable.  We are using C99 functions developed by others so I don't 
think we need to add tests for those.  If we have checked them informally, I 
think that is sufficient.  Then if someone does come across a problem, we can 
push it back to the C99 library people.

Regards,

Jerry