Subject: Re: Kernel library (libkern) Makefile glitch on pmax?
To: Jonathan Stone <email@example.com>
From: Chris G Demetriou <Chris_G_Demetriou@lagavulin.pdl.cs.cmu.edu>
Date: 06/27/1995 18:57:15
> >Not so 'odious,' of you realize that:
> > (1) one can't build 'kernel libraries' in a
> > machine-independent place without special
> > support for pulling in machine-specific options
> > (e.g. things dealing with global pointers, FP
> > regs, etc.), and,
> But it *used* to work, didn't it??
Iffy. for instance (unless i'm mistaken), not all of my alpha changes
made it back in...)
If you meant "did it do what it was trying to do, in the way it was
trying to do it," yes, it did. however, it was trying to do slightly
the wrong thing. 8-)
> > (2) having the libraries compile with different flags _per
> > kernel_ is a useful feature.
> More flags than profiled or not?? Well, if you say it's useful,
> it's useful.
yes, more flags than just "profiled." stuff like: optimization,
register (etc) allocation/use preferences, etc.
> I assumed someone was trying to clean up libkern, and use libc library
> functions, where appropriate (e.g., the fns are leaf fns, and
> originally derived from the same source), and didn't quite get it
Well, that may be what was going on, but i think the entire idea of
trying to do that is Just Plain Wrong, especially if it involves
removing the "redundant" sources from the kernel. the kernel _needs_
to be able to build w/o libc sources being present...