Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!linus!decvax!decwrl!pyramid!ut-sally!seismo!umcp-cs!chris
From: chris@umcp-cs.UUCP (Chris Torek)
Newsgroups: net.lang.c
Subject: Re: A diagnostic "gotcha"
Message-ID: <3181@umcp-cs.UUCP>
Date: Wed, 12-Feb-86 17:30:34 EST
Article-I.D.: umcp-cs.3181
Posted: Wed Feb 12 17:30:34 1986
Date-Received: Fri, 14-Feb-86 05:44:00 EST
References: <648@tymix.UUCP> <659@tymix.UUCP>
Organization: U of Maryland, Computer Science Dept., College Park, MD
Lines: 27

In article <659@tymix.UUCP> kanner@tymix.UUCP (Herb Kanner) writes:
>A line of C code that looked like:
>
>	node_mem[*dotp] = val & 0xff;
>
>produced the diagonistic "operands of + have incompatible types."  The
>error was that *dotp was not of type int, and a[b] being a euphemism for
>*(a + b), it complained about the operands of addition.

At a guess, I would say that `dotp' was declared as a pointer to
an enumerated type.  `char' and `short' are promoted to `int' in
addition expressions.  `enum's being what they are (half in and
half out of integer space at the moment), the 4.2 and 4.3 BSD C
compilers will not allow them to be added to pointers.  The ANSI
standard should clear this up.

In the meantime, a cast suffices to make the compiler accept the code:

	node_mem[(int) *dotp] = val & 0xff;

---though to my mind using enumerated types as indicies is `treading
on thin ice' (I have no explanation as to why I feel this way; I
just do).
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 1415)
UUCP:	seismo!umcp-cs!chris
CSNet:	chris@umcp-cs		ARPA:	chris@mimsy.umd.edu