CCL: Compiling Gaussian on EM64T machines
- From: "Perry E. Metzger"
<perry_+_piermont.com>
- Subject: CCL: Compiling Gaussian on EM64T machines
- Date: Tue, 18 Oct 2005 14:38:15 -0400
Sent to CCL by: "Perry E. Metzger" [perry(~)piermont.com]
"Grant Hill jgh105 _ york.ac.uk" <owner-chemistry**ccl.net>
writes:
> Sent to CCL by: Grant Hill [jgh105_._york.ac.uk]
> Hi all,
>
> I'm trying to compile Gaussian 03 (Rev C01) on a machine with dual
> EM64T intel Xeon cpus, running SuSE linux 9.3 (64-bit version, with
> all current updates applied). When I source the the g03.login script I
> get errors of the type:
>
> -#-: Badly formed number.
Okay, the first thing you need to do in debugging this is to get the
script to tell you what exactly is failing. I don't know offhand if
this is a csh or sh script you're having trouble with, but either way,
you should turn on debugging.
For both csh and sh, adding "-x" to the shell invocation will print
every command as it is executed, so you can determine which command
exactly it is. (I suspect that this is a csh script, but it either
way...)
You may have to edit the script to cause the #! line to include the -x.
> When I launch the build process I get similar errors throughout the
> log file suggesting to me that the shell (or perhaps automake /
> something similar) is having trouble. The errors are of the type:
>
> ./bsd/updatelink1 ../bsd/g03.make JUNK1=JUNK DO-LIB abt1dt.F
> -#-: Badly formed number.
> -f: Command not found.
> ar: No match.
> rm: No match.
> set noglob
> end
That looks like a shell getting very badly confused -- it is treating
switches as command names, and command names as globs. Again, doing
the "-x" trick will help in figuring out what command is going wrong
and thus causing everything to go out of wack.
Perry