Subject: Re: sysconf() et. al.
To: None <mark@aggregate.com, tech-kern@sun-lamp.cs.berkeley.edu, mike@cs.utah.edu>
From: Mark P. Gooderum <mark@aggregate.com>
List: tech-kern
Date: 03/16/1994 12:36:14
> These functions are all in 4.4-lite.  Its probably a toss-up whether you could
> implement them before lite comes out.  I would encourage you to wait unless
> you have an urgent need for them.

Acutally, I was just doing some quick checking and all the common values
can be reasonably gotten from compile time #defines as long as you are 
willing to accept that [f]pathconf() will assume ufs for now and not
get things right for odd exceptions like msdos fs's.

I did run accross some interesting descreprencies between some values
in <limits.h> and <sys/syslimits.h>.  

I'm inclined to do it since sysconf() is a basic POSIX function and having
it available will ease porting a lot of software.

I wouldn't be hurt if it was tossed out for the proper 4.4 lite solution
when it's integreated, but in all honesty, this is a ways off, no?

As it is now I could do it in about the same amount of time it would take
me to add the defines around porting one major package.

I only have a few questions which I'll follow up RSN.  Mainly a little
fuzzy on a what value a couple of values should have and whether it's 
a syscall or libfunc (and if #1 what numbers to use).

It doesn't need any kernel access per-se (assuming one blow's off a proper 
pathconf/fpathconf), but some of the #defines needed
are in <src/sys...> headers, but these *are* available during a scratch
build of the source tree by the time you actually build objects, right...or
not?

(It's been a looonnnggg time since I've built the whole tree from scratch
and my memory is fuzzy, etc.).

-Mark


------------------------------------------------------------------------------