>If you're making the claim that Sun's SPARC compiler always produces
>correct code, you're a pretty adventurous person.

As a side note, if you're building nvi, the reason v_txt.c is compiled
sans optimization is one version of Sun's compiler, still shipping last
I heard.

>> I am fine with that if it will buy us something... I am not convinced that
>> using __P is such a difficult thing.

>If that is the only ANSIsm we're talking about, then I'd have to
>agree.   Is that in fact true?

I was under the impression that there was some discussion, in the patch
whose discussion started this, of writing new-style definitions, rather
than old-style.  Unprotoize will obviously "correct" this, so I'm
inclined to think it's a win, in terms of simplicity of compilation.

Other examples of common differences:
	Do we use _P?  I dislike _P, and would like it to go away, because
	it's wasteful, but it's nearly-harmless.

	How does one write a function with variable arguments?

Both are trivially fixed with masses of preprocessor hacks.  Anyone wanna
spend three months eliminating the ritually unclean pre-ANSI code hacks,
and #defines, and see if it breaks anything?  I bet it would compile at
least a little faster - especially if we can simplify some of the very
commonly included files.

-s