From: Steve M. Robbins (steve@Nyongwa.CAM.ORG)
Date: 09/24/92


From: steve@Nyongwa.CAM.ORG (Steve M. Robbins)
Subject: Re: C News for Linux
Date: 25 Sep 1992 00:20:32 GMT

In article <L8oiRB1w164w@field.ichaos.nullnet.fi> kkk@field.ichaos.nullnet.fi (Kristo Kaarlo Matias) writes:
>
>I'm trying to "port" C News to Linux, and I'd like to know "best"
>answers for the following questions. (from 'build' script).

First of all, I should point out that someone has already put either binaries
or modified sources for ftp somewhere.

Second of all, I don't claim that the answers below are the best. Only that
they work for me (0.97 pl 2 or so).

><-------------------------------------------->
>C News has libraries for several kinds of Unix:
> bsd42 4.2BSD and successors
> usg AT&T System V
> v7 Version 7 (4.1BSD is pretty close, ditto Xenix)
> v8 Version 8, aka Eighth Edition
>Which best describes your system [v7]?

usg

>C News has libraries for small address spaces (16 bits) and big
>ones (preferably 32 bits, but anything rather bigger than 16).
>Which best describes your system [big]?

big
 
>Modern Unixes can generally use the setuid() system call to set the
>real and effective user ids to the current effective user id. In
[ ... ]
>(and it then invokes relaynews in return). Can this system do
>setuid(geteuidc)) to change the real uid/gid?

no

>Very old Unix systems needed the order of object modules in a library
>chosen very carefully. V7 introduced "ranlib" which removes the need
>for this. Recent System Vs have had the same facility built into "ar"
>(look for the "symdef" feature in the "ar" manual page) so "ranlib"
>Does your system use ranlib?

yes

>Does your "ar" have a "symdef" feature?

no
 
>Historically the only normal compilation option needed for most
>programs is -O, but again compilers, especially newer ones, differ.
>What options should be given to the compiler [-O]?

-On -- n can be any integer you like.

>The final linking ("ld") step of compiling might need an option,
>such as -n or -i, to produce the preferred form of executable file.
>On most modern systems the default is right. What options, if any,
>should be given for linking?

none
 
>C News often wants to ask how much disk space is available. The
>format of output from the "df" command unfortunately varies a lot,
>as does the availability of a system call to get the same information.
>C News knows about several different versions (the first three are
>preferred):
> statfs system with standard statfs() (SunOS, 4.4BSD, not System V)
> ustat system with ustat() (most System Vs)
> ultrix DEC Ultrix with DEC's own bizarre statfs()
> bsd 4.2/4.3BSD
> sysv old System Vs
> xenix some (all?) Xenixes; some System Vs, e.g. Microport, HP?
> sgi Silicon Graphics Iris systems
> v7 plain old style: no headers or fluff, just name and number
> null don't know or don't care how much space is available
>Which one is most appropriate?

For this, I'm not sure -- I'm using 'null'. If I was to experiment, I'd try
'ustat' first, then 'sysv'. I believe you need the GNU 'df' running for
'spacefor' to work right -- I didn't at the time I configured, which is why
I chose 'null'.

-- 
Steve Robbins  --  steve@nyongwa.cam.org
The World is rather sick and twisted.  Why should I be any different?