Newsgroups: news.software.b
Path: utzoo!henry
From: henry@utzoo.uucp (Henry Spencer)
Subject: Re: Cnews config file
Message-ID: <1989Aug11.215735.24616@utzoo.uucp>
Organization: U of Toronto Zoology
References: <1989Aug11.005854.12935@fxgrp.fx.com>
Date: Fri, 11 Aug 89 21:57:35 GMT

In article <1989Aug11.005854.12935@fxgrp.fx.com> grady@fxgrp.fx.com (Steven Grady) writes:
>Is there a reason why the config file (ie /usr/lib/news/bin/config normally)
>does not export the variable settings?   I've got alternate paths set up
>in the config file, but they don't get propagated (eg from rnews to newsspool)
>because the variables are not exported to the environment...

You're going about it the wrong way -- you want to start with the sources
and go through "build" again, which will generate a doit.bin that runs
"subst" to get the new variables into everything.  If you do get your new
variables down into newsspool via the environment, it will decide that
since they don't match the ones config.c knows about, you must be playing
tricks on it or running it for testing only... and it will renounce setuid
privileges to make sure nothing evil happens.

There is no terribly strong argument for or against exporting the variable
in config.  It simply doesn't make much difference.  All the shell files
pick it up explicitly anyway, and all the C programs rely on config.c as
the final authority (which means, as explained above, that you *can't*
change the variables by just changing config -- that only gets the shell
programs).  Nothing is likely to break if you do the exporting, but it
won't do what you think.
-- 
V7 /bin/mail source: 554 lines.|     Henry Spencer at U of Toronto Zoology
1989 X.400 specs: 2200+ pages. | uunet!attcan!utzoo!henry henry@zoo.toronto.edu