Subject: Re: interpic
From: Jim Rosinski (rosinski@bearmtn-e0.cgd.ucar.edu)
Date: Mon Jan 04 1999 - 16:35:30 MST
On Mon, 4 Jan 1999, Patrick D. Marquis wrote:
> I have been trying to compile the "interpic" tool for the CCM 3.6.6 and
> am running into difficulties. I am using Red Hat Linux 5.1 with the
> Portland Group's Fortran 90 compiler (pgf90).
[...]
> I ran into several problems along the way, including a double definition
> of the function "main", which I seem to have solved by renaming the main
> function in main.c to "MAIN_". In addition, I had to rename the Fortran
> 90 source files from .F90 to .f90 in order to get pgf90 to recognize the
> files.
That's one way to make things work. Another approach would be to modify the
file main.c to include the appropriate #define for C calling Fortran using
Portland Group's compilers. Note that the CCM code is not (yet) supported
under Linux. The Makefile for the main model includes a Linux option because
we (developers) happen to have access to a PC configured with pgf77.
Unfortunately, we don't currently have access to a machine with Portland
Group's f90 compiler, so we were unable to include similar mods for the CCM3
utility programs which require Fortran 90.
> The tool now compiles, with warnings (caused by pgf90 not allowing a
> real*16 type) but immediately core dumps with a segmentation fault when
> I attempt to run it with the example given on page 28 of the CCM3 users
> manual. (In the past, the segmentation fault has been caused by a too
> small stack size, but we have changed the stack to an unlimited size).
Under csh, the command "limit stacksize unlimited" does not truly render the
stacksize unlimited, but rather sets the "soft limit" to the "hard limit",
which is a finite number. I believe the hard limit is a boot time parameter
that you may have to increase if the code is still seg faulting.
Jim Rosinski
CCM Core Group
This archive was generated by hypermail 2b27 : Thu Jun 01 2000 - 09:26:25 MDT