Trying to find source of gfortran problems in Ubuntu

[FX Coudert]

> I already filed a bug report for Ubuntu's Fortran installation here:
>
> https://bugs.launchpad.net/ubuntu/+source/gcc-4.2/+bug/193299
>
> but I'm trying to get more insight on the source of the bugginess,  
> and I'm not even sure if it is a problem with GNU Fortran proper.  
> What stumps me is why I get NaNs when using the gfortran-4.2 that  
> comes from Ubuntu (see at above link), but not the binaries from  
> http://gcc.gnu.org/wiki/GFortranBinaries#GNU/Linux, or why I see  
> the same problem in OS X that I do in Ubuntu.

Really hard to be of any help without any details. What is the  
precise list of target, OS (including kernel and libc versions),  
compiler version (detailed, including revision number if available)  
and compilation options (e.g., -m32 and -m64) have you tested and  
what results do they give? From that, we can at least start thinking  
of a plan for you to track it down.

In particular, have you been able to reproduce it at all with a 4.3  
compiler? I understand that yes, you saw it on OS X with a 4.3  
compiler, but I want to make sure. Furthermore, "MacOS X Tiger XCode  
2.5" (gathered from you bugzilla report) is unspecific: is it PowerPC  
or Intel? 32-bit or 64-bit? What happens if you use the C 4.3  
compiler that comes with our gfortran package (/usr/local/gfortran/ 
bin/gcc)?

> Am I looking at problems with binutils, or a 64-bit problem, or what?

32/64 bit problem, why not. binutils, I rather doubt it. gcc/gfortran  
version mismatch, I seriously doubt it, because it would probably  
mean there is some wrong mixed-language code in Abinit.

FX

-- 
François-Xavier Coudert
http://www.homepages.ucl.ac.uk/~uccafco/