[2548] in Kerberos_V5_Development
Re: Prototype hell
daemon@ATHENA.MIT.EDU (E. Jay Berkenbilt)
Fri Oct 10 13:18:18 1997
Date: Fri, 10 Oct 1997 13:18:40 -0400
From: "E. Jay Berkenbilt" <ejb@ql.org>
To: sommerfeld@orchard.east-arlington.ma.us
Cc: marc@cygnus.com, tlyu@MIT.EDU, raeburn@cygnus.com, jhutz+@cmu.edu,
kenh@cmf.nrl.navy.mil, krbdev@MIT.EDU
In-Reply-To: <199710101556.PAA09247@orchard.east-arlington.ma.us> (message
from Bill Sommerfeld on Fri, 10 Oct 1997 11:56:36 -0400)
> I'd love to punt all the prototype crap.
I hope you mean the *non-prototype* crap.
> When I've suggested this in the past, Ted grew two new heads and
> told me to go back across the river :-)
I think it's time to revisit this decision.
It's been about five years since I've had to build something on a
platform which didn't have an ANSI compiler.
Right now, the burden is on people who have compilers that are
ANSI-compliant. (Hmm. I accidentally typed "complaint" at first, or
was that an accident?) At the very least, it seems time to shift the
burden over to people that do not have ANSI compilers. Ghostscript
historically has used a program called ansi2knr that basically can
used as a preprocessor to convert ANSI-style declarations to K&R-style
declarations. I'm not sure what else, if anything it does, or
whether it would be at all useful to include that or a reference to it
in the kerberos distribution.
Of course, ANSI-compliance gives you a lot more than just prototypes.
For example, string literals placed next to each other are
concatenated, you have use of the #elif preprocessor directive, etc.
Simply omitting support for compilers that don't support prototypes is
probably much less of a big deal than requiring a true ANSI-compliant
compiler. IMHO, the least (and possibly most) we should do is to stop
wasting energy supporting compilers that don't support prototypes. If
we want to explore options such as including ansi2knr, it could help;
otherwise, I agree with the sentiment that you lose if you don't have
an ANSI compiler in 1997.
