From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 13279 invoked by alias); 5 Jan 2007 21:06:52 -0000 Received: (qmail 13238 invoked by alias); 5 Jan 2007 21:06:43 -0000 Date: Fri, 05 Jan 2007 21:06:00 -0000 Message-ID: <20070105210643.13237.qmail@sourceware.org> X-Bugzilla-Reason: CC References: Subject: [Bug middle-end/30364] [4.1/4.2/4.3 Regression] Wrong variable ranges due to constant folding In-Reply-To: Reply-To: gcc-bugzilla@gcc.gnu.org To: gcc-bugs@gcc.gnu.org From: "gdr at integrable-solutions dot net" Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Id: List-Archive: List-Post: List-Help: Sender: gcc-bugs-owner@gcc.gnu.org X-SW-Source: 2007-01/txt/msg00329.txt.bz2 ------- Comment #8 from gdr at integrable-solutions dot net 2007-01-05 21:06 ------- Subject: Re: [4.1/4.2/4.3 Regression] Wrong variable ranges due to constant folding "jakub at gcc dot gnu dot org" writes: | Do the parenthesis matter in C? Yes, see paragraphes 11 dowward in the section 5.1.2.3. | They do matter in say Fortran, but in C I | think | (a - 20) + (b - 20) can be evaluated as (a + b) + (-20 + -20) or a - 20 - 20 + | b while that is permitted by the fortran standard, it is my understanding that *existing* practice has fortran compilers refrain from messing with parenthesis. However, this has become a middle-end issue; so I guess I must voice the concern for languages where unconditional re-association is not permitted. The C and C++ languages are example of those. -- Gaby -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30364