From: Bruce W Church <bwc1@cornell.edu>
Newsgroups: comp.parallel.pvm
Subject: Re: "wd=" does'nt work on Windows NT
Date: Wed, 24 Feb 1999 08:56:52 -0500
Organization: Cornell University
Sender: bwc1@cornell.edu (Verified)
Message-Id: <36D40524.639FD42F@cornell.edu>
References: <36D15EC9.FDBCD1ED@mechb.uni-stuttgart.de>
    <36D2BEE5.73C@pvmcube4.snu.ac.kr> <36D333D6.AFF6DF6E@cornell.edu>
    <36D3A66C.E5EBE185@mechb.uni-stuttgart.de>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Xref: ukc comp.parallel.pvm:8052


At the time I did try changing
PVM_TMP to the working directory, 
but then the pvm console could not 
communicate with the pvmd.  I have
not tried this on 3.4b6 so this problem 
may have been fixed.

-- Bruce 

Ralf Kuebler wrote:
> 
> Bruce W Church wrote:
> >
> > I haven't had a chance to test this
> > on the 3.4b6 version of pvm, but on an
> > older version for WIN-NT wd= was working properly
> > in the sense that spawned programs started
> > in the directory specified by wd=.  However
> > any calls to pvmlib routines (including
> > pvm_mytid() :-) ) changed the current
> > working directory to PVM_TMP.
> >
> > I worked around the problem by using fully
> > qualified filenames so I don't know if
> > this still occurs in 3.4b6.
> >
> 
> I want to start SIMULINK with PVM. SIMULINK
> looks for S-functions in the working directory.
> Consequently, it would be very inconvenient to
> use fully qualified filenames for all files.
> 
> Maybe one other solution is to set the PVM_TMP
> variable to the working directory manually
> before starting the application, or do you
> think there would be some unwanted side effects?
> And do you know how to set the PVM_TMP variable
> on Windows NT?
> 
> Ralf
> 
> --
> Ralf Kuebler, Institute B of Mechanics, University of Stuttgart
> Pfaffenwaldring 9, D-70550 Stuttgart, phone/fax: (+49)-711-685-6396/6400
> mailto:rk@mechb.uni-stuttgart.de
> http://www.uni-stuttgart.de/UNIuser/mechb/people/Kuebler/