Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ames!ncar!ico!ism780c!haddock!karl
From: karl@haddock.ima.isc.com (Karl Heuer)
Newsgroups: comp.lang.c
Subject: Re: address of function
Message-ID: <13751@haddock.ima.isc.com>
Date: 19 Jun 89 18:53:09 GMT
References: <2700@solo8.cs.vu.nl> <370@msor0.UUCP> <1817@auspex.auspex.com> <10361@socslgw.csl.sony.JUNET> <13731@haddock.ima.isc.com>
Reply-To: karl@haddock.ima.isc.com (Karl Heuer)
Organization: Interactive Systems, Boston
Lines: 26

>I can't think of any context in which the language cares whether an rvalue is
>a "constant" or not.

I must have been asleep when I wrote that.  (Actually, I misunderstood the
terminology as being yet another manifestation of the misbelief that "an array
is just a constant pointer", and failed to note that it really was relevant to
the question.)

Of course the language does care about constant expressions; they specify the
width of a bitfield, the value of an enumeration constant, the size of an
array, the label of a |case| statement, the conditional of a preprocessor #if,
and (in the example under discussion) the initializer for an object of static
storage duration.

The question was whether a cast is permitted in a constant expression.  The
answer is "Yes, but only from arithmetic types to integral types" in the first
four cases above (integral constant expressions); "No, not at all" in the case
of a preprocessor #if.  For initializers, the language seems to accept
anything that's well-defined, including pointer-to-pointer casts.

>>	int	this[100], *that = (int *)&this;

So this is indeed legal (though, as I said before, writing it as "&this[0]"
eliminates the need for the cast, and is almost certainly clearer).

Karl W. Z. Heuer (ima!haddock!karl or karl@haddock.isc.com), The Walking Lint