Subject: Re: values.h
To: Mauricio Tavares <raub@kushana.aero.ufl.edu>
From: J.T. Conklin <jtc@cygnus.com>
List: port-sun3
Date: 01/11/1996 10:53:23
>>>>> "Mauricio" == Mauricio Tavares <raub@kushana.aero.ufl.edu> writes:
Mauricio> I am planning on compiling scsiinfo in my 3/50.  Once I
Mauricio> tried to run its makefile, I got a few error messages:

I don't know anything about scsiinfo, but from your description it looks
very machine/OS dependent.

Mauricio> Without bothering about the sunos5.h messages for the time
Mauricio> being, we can see the compiler can't find values.h in
Mauricio> /usr/include.  

If I remember my history correctly, values.h is a header file that was
created somewhere on the SysIII/SysV branch of the UNIX family tree.
NetBSD doesn't have values.h and it is unlikely to be added because
most of the same constants (with different names) are available from
the standard header files <limits.h> and <float.h>.

Mauricio> Running find, I did find a values.h file in
Mauricio> /usr/include/g++.  Would it be equivalent to
Mauricio> /usr/include/values.h?

I think it might be close.  But note that the current version of
libg++ no longer provide values.h so it will disappear as soon as
libg++ is integrated into NetBSD.

Mauricio> If so, should I link it to /usr/include/values.h or simply
Mauricio> tell the compiler to look for header files in
Mauricio> /usr/include/g++?

If your program is really as machine dependent as I think it is, changing
it to use <limits.h> & <float.h> instead of <values.h> will be the least
of your worries.

	--jtc