<!doctype html public "-//w3c//dtd html 4.0 transitional//en">
<html>
All,
<p>Unfortunately this didn't seemed to fix things:
<p>setenv FC "f77 -i8"
<br>setenv F90 "f90 -i8"
<br>setenv FFLAGS "-i8"
<p>./configure --prefix=<dir> --with-mpi=<mpi?
<p>Still produced the same configure error:
<p>Checking whether size_t and int size match... no
<br>configure: error: This is a critical error; please report this size
mismatch to the developers!
<p>I'll keep tyring tommorow and let you know if I make any progress, does
anyone else have any other ideas?
<p>thanks for your help and quick response.
<p>--Brad
<p>Jianwei Li wrote:
<blockquote TYPE=CITE>Brad,
<p>I checked and found that Bill just said you can bypass the fortran
<br>interface since you don't need it.
<p>And here is the following answer from Tyce about how to bypass it:)
<p>Date: Thu, 21 Aug 2003 08:35:10 -0700
<br>From: Tyce Mclarty <mclarty3@llnl.gov>
<br>To: JB Gallagher <jbgallag@flash.uchicago.edu>,
<br>jianwei@ece.northwestern.edu,
<br> rross@mcs.anl.gov, parallel-netcdf@mcs.anl.gov,
<br> jbgallag@flash.uchicago.edu
<br>Subject: Re: Compiling parallel-netcdf 0.8.9 on QSC (ASCI Q)
<p>Can't you use the -i8 fortran flag to force 64 bit integers, so it will
<br>pass the configure test?
<p> From f90 man page:
<p> -i8 Makes
default integer and logical variables 8-bytes long
<br>(same
<br>
as the -integer_size 64 option). The default is
<br>-integer_size
<br>
32.
<p>Tyce
<p>Hope it will help you through.
<p>---
<p>Jianwei
<p>On Thu, 28 Aug 2003, JB Gallagher wrote:
<p>> All,
<br>>
<br>> I posted a question to the list a while ago about building the newest
<br>> parallel-netcdf on ASCI Q, I believe Bill Gropp got back to me on
my
<br>> problem, but I can no longer find his reply. I'm running into a problem
<br>> with the configure:
<br>>
<br>> checking whether size_t and int size match... no
<br>> configure: error: This is a critical error; please report this size
<br>> mismatch to the developers!
<br>>
<br>> And I believe Bill said something to the effect that this was put
in for
<br>> checks in building the fortran interface and implied that if I didn't
<br>> need the fortran interface (which I currently don't) I could bypass
this
<br>> check. Can anyone help me with this?
<br>>
<br>> After getting some mpi issues out of the way, I am able to get a
really
<br>> old version of parallel netcdf built on Q, but it doesn't work reliably
<br>> on greater than 8 processors.
<br>>
<br>> Thanks,
<br>>
<br>> --Brad
<br>>
<br>> --
<br>> Brad Gallagher
<br>> ASCI Flash Center
<br>> jbgallag@flash.uchicago.edu
<br>>
<br>>
<br>></blockquote>
<pre>--
Brad Gallagher
ASCI Flash Center
jbgallag@flash.uchicago.edu</pre>
</html>