🚀 go-pugleaf

In article <1992Dec15.142103.49328@news.th-darmstadt.de> wallmann@backus.pu.informatik.th-darmstadt.de (Natuerlich!) writes:
>What I expect to happen in expressions like these:   INT_MAX << 1
>is that the compiler behaves as if it were ultralame.

K&R defined C to be ultralame (as a replacement for assembly language),
except that optimizations could range from ultrlame to (sometimes) unsafe.
That's somewhat a different language from what we're dealing with now.

>It acts as if it doesn't precompute at compiletime but rather computes the
>result at runtime.

In both K&R-1 and in ANSI/ISO C, it doesn't matter if this gets precomputed
or not.  The execution of the program can't distinguish.  This entire
discussion has not been trying to distinguish such a point.

>So that in fact the default type for INT_MAX is signed int (since there
>isn't an U or an L anywhere) and the shifting operation operates on a
>signed int on that system.

INT_MAX is always signed int, for the reason you gave.  The word "default"
is meaningless here.  The shift operation operates on a signed int.

>If you want portability you write  (long) INT_MAX * 2 (or somesuch).

If you want portability you don't try to compute  INT_MAX << 1  in the
first place.  This is what the entire discussion has been about.

Incidentally, (long) INT_MAX * 2 also isn't portable.  For example, int
and long might both be 96 bits, and multiplying by 2 might overflow.
But in this discussion, no one's been trying to get a long out of this.
--
Norman Diamond       diamond@jit081.enet.dec.com
If this were the company's opinion, I wouldn't be allowed to post it.
"It's been a lovely recession."