Path: utzoo!attcan!uunet!lll-winken!lll-tis!ames!necntc!ima!haddock!karl
From: karl@haddock.ISC.COM (Karl Heuer)
Newsgroups: comp.lang.c++
Subject: Re: operator precedence question
Keywords: c++, precedence
Message-ID: <5341@haddock.ISC.COM>
Date: 20 Jul 88 15:31:09 GMT
References: <347@gt-eedsp.UUCP>
Reply-To: karl@haddock.ima.isc.com (Karl Heuer)
Distribution: na
Organization: Interactive Systems, Boston
Lines: 24

In article <347@gt-eedsp.UUCP> baud@gt-eedsp.UUCP (Kurt Baudendistel) writes:
>  for ( int i=0, t=s; i < 10; i++ ) ...
>[intending to perform "t=s" as an assignment, but instead finding it to be a
>declaration]

Basically, you've been hit by what could be considered a design flaw in the
syntax of the "for" statement.  In C, where the three components are
expressions, there's no ambiguity, and one can do multiple initializations via
the comma operator.  In C++, the first component is a statement instead of an
expression (note that all expressions are statements, but not vice versa); if,
as in this case, it's a non-expression statement, then the comma operator is
not legal.  Neither is a semicolon (normally the statement-equivalent of the
comma operator), because of the ambiguity this would create.

The appendix to Bjarne's C++ book suggests that
  for ({int i=0; t=s;} i < 10; ++i) ...
should work, but I haven't tested it.  An even kludgier approach is to use
  for (int i=(t=s, 0); i < 10; ++i) ...
which should be a legal use of the comma operator.

Personally, I disrecommend using a declaration as the first component of a for
statement, since it (improperly, in my opinion) extends to the end of the
enclosing block instead of being restricted to the for statement itself.

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