From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 4570 invoked by alias); 8 Jul 2004 19:28:10 -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 4550 invoked by uid 48); 8 Jul 2004 19:28:09 -0000 Date: Thu, 08 Jul 2004 19:28:00 -0000 Message-ID: <20040708192809.4549.qmail@sourceware.org> From: "ebotcazou at gcc dot gnu dot org" To: gcc-bugs@gcc.gnu.org In-Reply-To: <20040621084809.16109.pluto@pld-linux.org> References: <20040621084809.16109.pluto@pld-linux.org> Reply-To: gcc-bugzilla@gcc.gnu.org Subject: [Bug target/16109] error: architecture mismatch on "fzero". X-Bugzilla-Reason: CC X-SW-Source: 2004-07/txt/msg00826.txt.bz2 List-Id: ------- Additional Comments From ebotcazou at gcc dot gnu dot org 2004-07-08 19:28 ------- > If the compiler is configured with --with-cpu=v7, then it defaults to -m32. This one I've always found strange, but now I see how it works. It's different on Solaris, because sol2-64.h unconditionally sets MASK_64BIT in TARGET_DEFAULT. > #define OPTION_DEFAULT_SPECS \ > {"cpu", "%{!m64:%{!mcpu=*:-mcpu=%(VALUE)}}" }, \ > ... > means that -mcpu=v7 will be passed if no -mcpu is given AND -m64 is not > present on the command line. If -m64 is present and -mcpu is not given, > CC1_SPEC will add -mcpu=ultrasparc. > Similarly for a compiler defaulting to -m64. Aaaaah! I should have grasped that when you posted the OPTION_DEFAULT_SPECS patch for Linux. We have the "-m64 doesn't shut down -mcpu=v7" problem on Solaris in 3.4.x (PR target/16416) so I'm going to shamelessly steal your patch. Thanks for the explanation. [About the original problem:] It seems to me that the flags are missing on the line: sparc64-pld-linux-gcc ../sysdeps/sparc/sparc64/memset.S [...] -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16109