NetBSD Problem Report #42157

From dan@mcmahill.net  Tue Oct  6 17:43:26 2009
Return-Path: <dan@mcmahill.net>
Received: from mail.netbsd.org (mail.netbsd.org [204.152.190.11])
	by www.NetBSD.org (Postfix) with ESMTP id E76C763B8B6
	for <gnats-bugs@gnats.netbsd.org>; Tue,  6 Oct 2009 17:43:26 +0000 (UTC)
Message-Id: <20091006174323.692E9246027E@bondage.mcmahill.net>
Date: Tue,  6 Oct 2009 13:43:23 -0400 (EDT)
From: dmcmahill@NetBSD.org
Reply-To: dmcmahill@NetBSD.org
To: gnats-bugs@gnats.NetBSD.org
Subject: math/lapack versus gfortran
X-Send-Pr-Version: 3.95

>Number:         42157
>Category:       pkg
>Synopsis:       math/lapack needs gfortran fixes
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    asau
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Oct 06 17:45:00 +0000 2009
>Last-Modified:  Tue Oct 06 18:54:10 +0000 2009
>Originator:     Dan McMahill
>Release:        NetBSD 5.0_STABLE
>Organization:
NetBSD

>Environment:


System: NetBSD pkghacker 5.0_STABLE NetBSD 5.0_STABLE (PKGHACKER) #1: Sat Jun 20 20:42:58 EDT 2009 root@pkghacker:/usr/cvs/src/sys/arch/i386/compile/PKGHACKER i386
Architecture: i386
Machine: i386
>Description:

math/lapack has 3 minor issues which need to be addressed to be compatible with gfortran (available in lang/gcc44).
As an aside, we need math/lapack to work with gfortran because some programs like scilab require gfortran for their
newer releases.

1)  we now need to use the intrinsic instead of external etime().

In patches/patch-ac we'd change EXT_ETIME to INT_ETIME.  However, this may need to be a switch
based on the actual fortran compiler used.  

2)  early on in the build on i386, the single precision tests hang.  Adding -ffloat-store
to FFLAGS in math/lapack/Makefile like:

FFLAGS+=		-ffloat-store

gets this to work again, but this is both i386 specific and gcc specific.

3) I didn't have a problem with this, but I noticed that patch-ad has:

.f.o:
       ${LIBTOOL} --mode=compile --tag=CC ${FC} ${FFLAGS} -c $<


in particular note that we are using the CC tag to libtool with *fortran*.  Shouldn't that be
the F77 tag?  Of course that assumes we even had fortran installed properly before libtool which is
a bit of a chicken and egg issue.


>How-To-Repeat:

/etc/mk.conf:

CC=/opt/gcc44/gcc44/bin/gcc
CXX=/opt/gcc44/gcc44/bin/g++
FC=/opt/gcc44/gcc44/bin/gfortran

I built the gcc44 package into PREFIX=/opt/gcc44 and then pointed to this for my pkgsrc builds.


>Fix:

Figure out how to automatically use the old behavior versus the patched code here depending on the
fortran compiler in use.



>Release-Note:

>Audit-Trail:

Responsible-Changed-From-To: pkg-manager->asau
Responsible-Changed-By: asau@NetBSD.org
Responsible-Changed-When: Tue, 06 Oct 2009 18:54:10 +0000
Responsible-Changed-Why:
Take (working on it anyway).


>Unformatted:


 pkgsrc as of 2009-10-06

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.39 2013/11/01 18:47:49 spz Exp $
$NetBSD: gnats_config.sh,v 1.8 2006/05/07 09:23:38 tsutsui Exp $
Copyright © 1994-2007 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.