pnetcdf tests

[Jim Edwards]

Jianwei,

Thanks, that seems to have solved it. 

On Mon, 2005-06-06 at 12:00 -0500, Jianwei Li wrote:
> Hi Jim,
> 
> You are right that nc_test/nf_test are designed for single process test.
> There are two steps to run nc_test (nf_test is similar):
>  	(1) create test.nc by 	"<mpirun -np 1> ./nc_test -c"
>  	(2) run the test by 	"<mpirun -np 1> ./nc_test -v"
> I will make sure that these lines of documentation DO appear in README
> in next release!
> 
> As to the FORTRAN style comments, I'm sure I addressed that issue long
> ago explicitly for AIX platform. It probably got reverted accidentally
> in later releases. We'll fix that again for sure!
> 
> Thanks,
> Jianwei
> 
> On Mon, 6 Jun 2005, Jim Edwards wrote:
> 
> > Hi, I have just installed parallel-netcdf-1.0.0-pre2 on two platforms
> > 1. An AIX 5.1 system with OBJECT_MODE=64
> > 2. A BGL platform (basically an ppc linux machine)
> >
> > Both platforms use IBM's xlc/xlf compilers.  I couldn't find any
> > documentation about the tests, but after a little experimenting I
> > decided that they should be run from just one processor - is that
> > correct?  On both platforms nc_test passes everything but nf_test
> > fails a large number of tests with unexpected results.  Before I could
> > compile the fortran tests I had to modify the .inc files and add ! in
> > front of c-style comments.
> >
> > Any ideas on why i get all the unexpected results?
> >
> > -- 
> > Jim Edwards
> > IBM Applications Analyst
> > National Centers for Atmospheric Research
> > Boulder CO 80305     303-497-1842
> >
-- 
Jim Edwards
IBM Applications Analyst
National Centers for Atmospheric Research
Boulder CO 80305     303-497-1842