From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 12150 invoked by alias); 12 Apr 2005 14:46:46 -0000 Mailing-List: contact gcc-bugs-help@gcc.gnu.org; run by ezmlm Precedence: bulk List-Archive: List-Post: List-Help: Sender: gcc-bugs-owner@gcc.gnu.org Received: (qmail 12096 invoked by uid 48); 12 Apr 2005 14:46:40 -0000 Date: Tue, 12 Apr 2005 14:46:00 -0000 Message-ID: <20050412144640.12095.qmail@sourceware.org> From: "tow21 at cam dot ac dot uk" To: gcc-bugs@gcc.gnu.org In-Reply-To: <20050216120130.20006.tow21@cam.ac.uk> References: <20050216120130.20006.tow21@cam.ac.uk> Reply-To: gcc-bugzilla@gcc.gnu.org Subject: [Bug fortran/20006] $ format extension doesn't work X-Bugzilla-Reason: CC X-SW-Source: 2005-04/txt/msg01562.txt.bz2 List-Id: ------- Additional Comments From tow21 at cam dot ac dot uk 2005-04-12 14:46 ------- (In reply to comment #6) > Fatal I/O error for an extension that everybody agrees on? Wahoo! I'm not > > sure any compiler does that... But "everyone" doesn't agree on it. Otherwise it would be standard, not an extension,. If I compile with -pedantic -std=f95, then I'm saying I don't want any extensions, regardless of how many people agree on them. > In my mind, being pedantic (that is, issuing fatal errors) makes sense at > compile-time, but it is very harsh to do so at runtime. I can imagine how, > as a researcher, I would compile with -pedantic, be glad that no error turn > out, then run a one-week simulation code and see it issue a fatal error > instead of printing the results at the end... Then I think you're putting too much trust in -pedantic. This is why, for example, even if my codes compile with -pedantic, I'll run them thoroughly with -C. You can't catch every possible error at compile time. I would be equally annoyed if having run everything through with full -pedantic and error-checking on gfortran, I assumed that my code was fully standard-conforming, moved my code across to NAG fortran, and it crashed with Invalid edit descriptor beginning with '$' Program terminated by fatal I/O error zsh: 27142 abort (core dumped) ./a.out which it would, because that's what it does now. Not, of course, that we can catch every standard violation, but I would expect that this be caught. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20006