SUMMARY on G94-Linux running problem
Dear CCLers,
last week I submitted the following mail:
-------------------------------------------------------
>I am trying to use Gaussian 94 on a PC Pro 200 operating with Linux but it
>does not work.
>My system is a Pentium Pro 200 on a Venus Intel Motherboard, cache 256 KB,
>RAM 64 MB, 2 HDs EIDE.
>I use Linux 2.0.0, gcc 2.7.2 version, f2c 19950110 version.>
>
>The program is compiled well, there are only a lot of Warning messages like
>these:
>
>Warning on line 26 of bdump.f: non-character datum nw initialized with
>character string
>Warning on line 49 of bufrwf.f: inconsistent calling sequences for fileio,
>Warning on line 309 of dout2e.f: local variable zero never used
>Warning on line 70 of dset2e.f: local variable maxnwb never used
>
>and so on.
>
>When I tried to run some tests, the program begun but after a short time the
>program stopped without messages of error, in the working directory there is
>a core file and in the log file there is the following message:
>
>Leave Link 101 at Tue Mar 11 15:42:21 1997, MaxMem= 4000000 cpu:
0.0
>
>I tried to test the RAM memory with different tests, but apparently it works
>well.
>Has anyone any suggestions ? Furthermore, any Gaussian user with Linux
>operating system could tell me wich version of Linux, gcc and f2c does it
>work well ?
-------------------------------------------------
I thank all the CCLers who have send me suggestions, in particular I have to
thank dr. Douglas J. Fox, director of Gaussian Technical Support, because he
gave me the solution of the problem. Following his instruction this is the
solution.
There is a problem with the libc.so.XXXX libraries of Linux.
I have libc.5.3.12 but it doesn't work and Dr. Douglas told me that even
libc.so.5.4.7 has the same problem.
libc.so.5.2.18 works fine and it is possible downloading it from every linux
mirror, for istance I took it from:
ftp://sunsite.cnlab-switch.ch/mirror/linux/GCC
After that, you have to put the library in a directory like /lib and before
starting G94 you have to give the following instruction:
setenv LD_PRELOAD /usr/local/lib/libc.so.5.2.18
from csh shell, I have not found the equivalent instruction from bsh!
It is possible to eliminate the warning messages by using g77 after some
modifications in f2c.make, as Bernd Melchers has suggested.
At the end of this message I will put the messages that I received, I think
that there is other interesting comments and suggestions.
Thank you again.
Bye
Giuseppe Ermondi
-------------------------------------------------------------
-------------------------------------------------------------
In-Reply-To: <199703112326.SAA27760 "-at-" www.ccl.net> from
"Giuseppe" at Mar
11, 97 06:26:07 pm
Hi Giuseppe,
Do you use RedHat 4.0? I had similar problems, and they vanished
when I downgraded to RedHat 3.03. I don't now where the problem lies
exactly, but seems that some RH 4.0 dynamic libraries are somewhat buggy -
the G94 compiled on RH 3.0 (and linked statically) runs fine under RH 4.0.
I have not had time to test new RedHat 4.1 yet.
If you have more questions please don't hestitate to contact me.
Regards,
Peeter
--
Peeter Burk Jakobi 2, EE2400 Tartu, Estonia
Institute of Chemical Physics Phone (372-7) 465-265
Tartu University Fax (372-7) 465-264
Estonia E-mail peeter "-at-" chem.ut.ee
----------------------------------------------
----------------------------------------------
Giuseppe,
...
> and so on.
>
> When I tried to run some tests, the program begun but after a short time
the
> program stopped without messages of error, in the working directory there
is
> a core file and in the log file there is the following message:
>
> Leave Link 101 at Tue Mar 11 15:42:21 1997, MaxMem= 4000000 cpu:
0.0
>
> I tried to test the RAM memory with different tests, but apparently it
works
> well.
> Has anyone any suggestions ? Furthermore, any Gaussian user with Linux
> operating system could tell me wich version of Linux, gcc and f2c does it
> work well ?
>
> If my problem was not of general interest please send me an E-Mail to
> ermondi "-at-" ch.unito.it.
>
I just compiled my own program (molden) on to different linux2.* versions
with f2c and g77, the compilation was ok but the g77 versions core dumps
immediately in a system routine __crt_dummy called from main. The f2c
version runs once or twice but then might crash a third time.
These same executables run fine on linux1.2.13. So it is my guess
that there is an operating system bug in linux2.* which manifest it self
only with relatively large programs, a memory management bug I guess.
Regards, Gijs
--
+----------------------------+-----------------------------------+
Gijs Schaftenaar, Drs. | CAOS/CAMM Center
Email: schaft "-at-" caos.kun.nl | University of Nijmegen
URL : http://www.caos.kun.nl/staff/schaft.html
Tel. : +31 24 3653369 | Toernooiveld 1
Fax : +31 24 3652977 | 6525 ED Nijmegen, The Netherlands
+-------- CAOS/CAMM is the Dutch National Node in EMBnet --------+
-------------------------------------------------------------
-------------------------------------------------------------
X-Sender: wagenert "-at-" pprz03.HRZ.Uni-Marburg.DE
In-Reply-To: <199703112326.SAA27760 "-at-" www.ccl.net>
On Tue, 11 Mar 1997, Giuseppe wrote:
Dear Giuseppe,
>
> I am trying to use Gaussian 94 on a PC Pro 200 operating with Linux but it
> does not work.
> My system is a Pentium Pro 200 on a Venus Intel Motherboard, cache 256 KB,
> RAM 64 MB, 2 HDs EIDE.
> I use Linux 2.0.0, gcc 2.7.2 version, f2c 19950110 version.
>
> The program is compiled well, there are only a lot of Warning messages like
> these:
[...]
> Has anyone any suggestions ? Furthermore, any Gaussian user with Linux
> operating system could tell me wich version of Linux, gcc and f2c does it
> work well ?
>
> If my problem was not of general interest please send me an E-Mail to
> ermondi "-at-" ch.unito.it.
we are running GAUSSIAN on a PPro 200 box under Linux with no problems at
all. I had the same trouble compiling the code for the first time because
of some problem with the f2c library which I had to fix manually. I was
somewhat disturbed by the amount of errors and warnings the compiler gave
(about 1200 as far as I remember) and the performance of the program so I
decided to try the native fortran compiler g77 (version 0.59.18 I think).
The compilation went with hardly any warnings and the code produced is
much faster compared to the f2c translation.
Manipulating the Makefile is fairly easy: you simply have to enter the g77
as the Fortran compiler, comment out the f2c lines and give the right
files (and file endings) in the compiler line (say, changing *.c to *.f).
If you have trouble getting things to work send me a mail and I will see
if I can find the MakefileFrom chemie.fu-berlin.de!melchers Wed Mar 12 11:13:55
1997
Message-Id: <m0w4l2h-008peJC "-at-" Bose.Chemie.FU-Berlin.DE>
From: melchers "-at-" chemie.fu-berlin.de (Bernd Melchers)
Subject: Re: CCL:G:G94 on Linux: running problem
To: ermondi "-at-" silver.ch.unito.it (Giuseppe)
Date: Wed, 12 Mar 1997 11:13:55 +0100 (MET)
In-Reply-To: <199703112326.SAA27760 "-at-" www.ccl.net> from
"Giuseppe" at Mar
11, 97 06:26:07 pm
X-Mailer: ELM [version 2.4 PL24 ME8d]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 701
Status: OR
'Giuseppe wrote:'
>
>
> Dear CCLers,
>
> I am trying to use Gaussian 94 on a PC Pro 200 operating with Linux but it
> does not work.
> My system is a Pentium Pro 200 on a Venus Intel Motherboard, cache 256 KB,
> RAM 64 MB, 2 HDs EIDE.
> I use Linux 2.0.0, gcc 2.7.2 version, f2c 19950110 version.
Why do you not use the real gnu fortran compiler, g77, instead of f2c.
I though, f2c is outdated.
Probably, you would have much less problems.
Ciao
Bernd
--
Bernd Melchers | melchers "-at-" FU-Berlin.DE
Freie Universitaet Berlin | "We don't write software, we compose it."
AG Macromelecular Modelling - Prof. Dr. E.W. Knapp
for more information see http://www.chemie.fu-berlin.de/~melchers
-------------------------------------------------------------
-------------------------------------------------------------
X-Mailer: ELM [version 2.4 PL23]
Hi there !
I have had exactly the same problems as you with exactly
the same machine and linux version.
I spent alot of time trying to sort this problem out
and finally decided to get the new version of g94 - version e.1 -
which compiles and works fine. The only problem with it
is that I had to put all the full paths of the libraries
in the make files for lm and lf2c.
Good luck !
Phil Sinclair.
--
##############################################################################
Phillip Sinclair
Learn from yesterday, Royal Institution of Great Britain
Live for today, 21 Albemarle Steet
Look to tomorrow, LONDON W1X 4BS, UK
Rest this afternoon. Snoopy. Phone: +44 (0)171-409-2992
email: phillip "-at-" ri.ac.uk
##############################################################################
----------------------------------------------------------
----------------------------------------------------------
X-Sender: smolnar "-at-" pop.service.ohio-state.edu
X-Mailer: Windows Eudora Pro Version 2.2 (32)
X-Priority: 1 (Highest)
I would pay attention to the warnings about inconsistent calling sequences.
It's my understanding that the particular compiler that you're using, the
same one that I've used to compile MOPAC7 and GAMESS for my platform, is
apparently the only one in the world that pays strict adherance to the
fortran standard with regard to the placement of source code statements.
I'm sorry that I can't be more percise, but I'm only parroting what I've
been told by a chap at the University of Minnesota who seems to know what
he's talking about.
I'd very much like to try compiling Gaussian 94 on my platform, but I don't
have the source code.
Hope this has been of some help.
Good luck.
Steve Molnar
______________________________________________________________________________
Stephen P. Molnar, Ph. D. Life is a fuzzy set
Foundation for Chemistry multivariant and stochastic
Technology Consultants
614.486.4076
----------------------------------------------------------
----------------------------------------------------------
In-Reply-To: <199703112326.SAA27760 "-at-" www.ccl.net> from
"Giuseppe" at Mar
11, 97 06:26:07 pm
X-Mailer: ELM [version 2.4 PL24]
Dr. Ermondi,
You do not note which version of libc you use but I suspect that is the
problem. I know for a fact that libc.so.5.4.7 displays this problem but
earlier versions like libc.so.5.2.18 and most likely libc.5.3.12 do not.
The first is the one I use but libc.5.4.7 came with my Slackware distribution.
You can use this preferentially with Gaussian by using the environment
variable LD_PRELOAD,i.e.
setenv LD_PRELOAD /usr/local/lib/libc.so.5.2.18
before running G94. I simply roll this into a small script or set up an
alias which sets this before running G94.
Douglas J. Fox
Director of Technical Support
help "-at-" gaussian.com
----------------------------------------------------------------
----------------------------------------------------------------
--
...................................................................
\\\|/// Giuseppe Ermondi
| | Dipartimento di Scienza e Tecnologia del Farmaco
|O^O| Via P. Giuria, 9
( (_) ) I-10125 Torino - ITALIA
\ o /
~V~ Tel: +39-11-6707680 FAX: +39-11-6707687
/ \ E-mail: ermondi "-at-" ch.unito.it
...................................................................