Re: Oskit-mach success with gcc-3.0

[Roland McGrath]

> At lunch I started a compile of oskit with gcc-3, and it appears to
> have built.  I'm going to play with an all gcc-3 based
> oskit+oskit-mach when I get home tonight.

Ok.  I don't know what to expect from this.  The problems of compiler vs
linux drivers tend to be of the nature that an individual driver won't work
right in some particular circumstance, so a lot of different testing could
be required to actually know there aren't problems.  The current oskit
linux code is from 2.2.12, so you can probably find some information about
what to expect from compiling linux-2.2.12 with gcc3 and that may apply.

> This patch to Makefile.in was reported by James Morrison a few months
> ago:

This should not be required.  My build is fine without it.  

The compiler by itself with the CC and CFLAGS settings you put in the
environment (or don't) for configure need to be correct for finding the
oskit headers and libraries already.  In my case, I have a i686-gnu cross
compiler that has oskit installed in its .../i686-gnu/{include,lib} (just
use make install prefix=.../i686-gnu to install oskit).  If you are using a
native compiler on an x86 GNU/Linux or *BSD system, then the default
installation of oskit in /usr/local ought to be found by default.

As to the removal of -loskit_smp, that obviously breaks the SMP build.
Having the library there when building without SMP is harmless.

I can't imagine a reason for these changes that does not involve a broken
compiler installation or an incorrect parameterization of configure.
If there is a problem, please show me fresh details.