From owner-freebsd-ia64@FreeBSD.ORG Fri Sep 22 04:16:10 2006 Return-Path: X-Original-To: ia64@freebsd.org Delivered-To: freebsd-ia64@FreeBSD.ORG Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id D6A7C16A501 for ; Fri, 22 Sep 2006 04:16:10 +0000 (UTC) (envelope-from mi+kde@aldan.algebra.com) Received: from aldan.algebra.com (aldan.algebra.com [216.254.65.224]) by mx1.FreeBSD.org (Postfix) with ESMTP id 2499B43D68 for ; Fri, 22 Sep 2006 04:16:08 +0000 (GMT) (envelope-from mi+kde@aldan.algebra.com) Received: from aldan.algebra.com (aldan [127.0.0.1]) by aldan.algebra.com (8.13.7/8.13.7) with ESMTP id k8M4G8mB051522 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-SHA bits=256 verify=NO) for ; Fri, 22 Sep 2006 00:16:08 -0400 (EDT) (envelope-from mi+kde@aldan.algebra.com) Received: from localhost (localhost [[UNIX: localhost]]) by aldan.algebra.com (8.13.7/8.13.7/Submit) id k8M4G7wB051521 for ia64@freebsd.org; Fri, 22 Sep 2006 00:16:07 -0400 (EDT) (envelope-from mi+kde@aldan.algebra.com) From: Mikhail Teterin To: ia64@freebsd.org Date: Fri, 22 Sep 2006 00:16:07 -0400 User-Agent: KMail/1.9.1 X-Face: %UW#n0|w>ydeGt/b@1-.UFP=K^~-:0f#O:D7whJ5G_<5143Bb3kOIs9XpX+"V+~$adGP:J|SLieM31VIhqXeLBli" Cc: Subject: troubles with devel/icu on ia64 X-BeenThere: freebsd-ia64@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Porting FreeBSD to the IA-64 List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Fri, 22 Sep 2006 04:16:10 -0000 Hello! As you may know, the devel/icu port does not build on FreeBSD/ia64. Or, rather, it builds, but crashes while running the vendor's self-tests. The crashes tend to occur in different places depending on the compilation flags, but are still very reproducible. Even with `-O0 -g'. The developers (from IBM) dismiss this as a "compiler bug". However, all our other architectures (including the fickle ones like sparc64) have no problems -- if it is a compiler bug, it must ia64-specific AND not affect any of other code... Although few ports run self tests post-build, the ones that do have no problems. The few problems reported from users of i386 were traced to overly aggressive compiler flags (such as `-O2 -funroll-loops'). I tried using Purify on the test binaries (on Solaris/sparc), but even after fixing the few Purify-reported buglets the crashes are still here. What can possibly be wrong with FreeBSD/ia64? The library (and the tests) use threads aplenty... Would anyone care to investigate? The modified port can be found on pluto1.freebsd.org:/p/scratch/mi/icu-36. Thanks! -mi