Dear NCAR Graphics users,
We have been building and installing NCAR Graphics 4.3.1 on a HP Proliant
DL380 running RedHat Linux 8.0, with separate builds performed for the pgf90
and ifc 7.1 compilers. Along the way we have encountered problems with
creating the 'ictrans' binary required by 'idt' to view gmeta files. Two of
my colleagues have inspected the source code and text output from the
makefiles, and have traced the problem to an unsuccessful link to f_cpystr
(see the attached e-mail below). Does anyone know of a solution to this
problem?
Thanks,
Eric Kemp
eric.kemp@ngc.com
> -----Original Message-----
> From: Tonkin, Martha E.
> Sent: Wednesday, July 16, 2003 12:33 PM
> To: Kemp, Eric M.
> Cc: Alliss, Randall J.; Applegate, Sheldon L.; Link, Robert P.
> Subject: Problem Building NCAR's ictrans on Mars
>
> Eric,
>
> As I told you, the build for the ictrans binary cannot resolve a reference
> to f_cpystr in
> the loader/link phase. The missing function is apparently a C
> implementation of FORTRAN
> code that copies one string to another. The NCAR function that wants
> f_cpystr is found in
> file fort_ictrans.f , with subroutine name GIARGS. The object module
> from the compile
> of fort_ictrans.f is stored in quasi library libictrans.o, which is linked
> with main.o in the ictrans
> subdirectory to create the ictrans executable.
>
> I'm attaching the output of the make commands for both libictrans.o and
> for ictrans.
> Note that these particular instances were the build using the Intel
> FORTRAN compiler,
> but the same type thing happens with the Portland Group(PG) FORTRAN
> compiler, which
> isn't surprising as it appears to be related to the gcc compiler/loader.
> (Note that
> we are using gcc version 3.2, on RedHat Linux 8 with the 2.4.20-13.8smp
> kernel.
> The Intel FORTRAN compiler is Version 7.1 Build 20030307Z ).
>
> Also note that I wrote a test program which I'm attaching, that mimics
> the
> string copies performed in GIARG. The test program compiled and linked
> properly with both Intel and PG FORTRAN compilers.
>
> I hope this is clear; I will be out tomorrow (Thursday) but will return
> Friday. Sheldon
> looked at this with me, as I thought he might see something I missed, and
> he
> probably can answer any questions you might have if I'm not around.
>
> Martha
>
> <<giargs.f>> <<make_ictrans.out>> <<make_libictrans.out>>
> <<test_char.f>>
>
>
>
>
This archive was generated by hypermail 2b29 : Mon Jul 21 2003 - 07:22:36 MDT