20040402: netcdf on Linux+pgf90, f90/test failure

Natalie,

>Date: Fri, 2 Apr 2004 13:10:38 -0800 (PST)
>From: Natalie Perlin <address@hidden>
>Organization: Oregon State University
>To: address@hidden
>Subject: Re: 20040402: netcdf on Linux+pgf90, f90/test failure 
>Keywords: 200404022013.i32KDHCT021202

The above message contained the following:

> Thank you, Steve, a lot for the prompt reply!!!
>   
>   Also, as the test printout (included in the body of the message)
> reported differences in char between the two files, is it the error
> caused by the bug in the compiler?..

Yes.

> It would be extremely useful to
> get some kind of the idea of this bug, as we have some othe troubles
> when using pgf90 compiler for our parallel runs with MPICH, but do
> not know whether it is the compiler problem of some other things (it
> works for a single node, but the blows up during multi-node run). It
> might be too specific information and details to explain, but I'm
> just wondering if you may say few more words about the pgf90 compiler
> bug? Please, disregard this request if it's complicated,

The relevant lines in the Fortran-90 test program (f90/netcdf_test.f90)
are the following:

  call check(nf90_put_var(ncFileID, pressVarID, pressure(:, :, 1)) )
  call check(nf90_put_var(ncFileID, pressVarID, pressure(:, :, 2), start = (/ 1,
 1, 2 /)) )

The first call to nf90_put_var() works correctly, but the "pressure(:,
:, 2)" expression in the second call obtains an array with wrong
dimensionality (it's 2-D instead of 3-D) and wrong starting point (it's
at the beginning of the "pressure" array instead of halfway through
it -- which is why the output values are wrong).

>   Thanks once more,
>   Natalie.

Regards,
Steve Emmerson