>At 10:07 am 8/3/96, Peter Seebach wrote:
>That is exactly what I was trying to say, only clearer.  If we can port gcc
>to a platform, we have an ANSI compiler.  If we can't even port gcc, how
>are we going to get anywhere?  In fact, is there any reason why *any* of
>the source tree *has* to compile with any compiler apart from gcc, since
>any working NetBSD installation has gcc, and all platforms currently (and
>hopefully future) supported have a gcc?  OK, it might be neat not to depend
>on gcc, but I am willing to bet we do in our Makefiles.  In fact, we depend
>on berkeley make, so why not depend on gcc.  Do other compilers understand
>-Werror, for example?

I think it's sane to use, for instance, -Werror, because we can easily fix
it if we switch compilers.  Assume that we discover that lcc now supports
every platform we want, compiles faster, and optimizes better.  (Well, one
out of three ain't half bad.  It's two thirds bad.)  If we're reasonably
careful about not using extensions, we can switch easily.  I disagree
with the gcc manual's claim that there is less and less reason to use other
compilers; as ANSI becomes more and more widespread, there is less and less
reason to violate the standard.

I would like to see NetBSD continue to become a more and more
standards-friendly OS.  It's nice having documentation on which
things are ANSI, which are POSIX, and which are BSD.  This kind of
thing makes NetBSD a good choice for development work.

-s