From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gcc-return-117482-listarch-gcc=gcc.gnu.org@gcc.gnu.org>
Received: (qmail 23263 invoked by alias); 19 Jul 2005 20:14:12 -0000
Mailing-List: contact gcc-help@gcc.gnu.org; run by ezmlm
Precedence: bulk
List-Archive: <http://gcc.gnu.org/ml/gcc/>
List-Post: <mailto:gcc@gcc.gnu.org>
List-Help: <http://gcc.gnu.org/ml/>
Sender: gcc-owner@gcc.gnu.org
Received: (qmail 22921 invoked by uid 22791); 19 Jul 2005 20:14:06 -0000
Received: from c-24-61-23-223.hsd1.ma.comcast.net (HELO cgf.cx) (24.61.23.223)
    by sourceware.org (qpsmtpd/0.30-dev) with ESMTP; Tue, 19 Jul 2005 20:14:06 +0000
Received: by cgf.cx (Postfix, from userid 201)
	id B86A413C261; Tue, 19 Jul 2005 16:14:04 -0400 (EDT)
Date: Tue, 19 Jul 2005 20:14:00 -0000
From: Christopher Faylor <me@cgf.cx>
To: gcc@gcc.gnu.org, dannysmith@users.sourceforge.net,
	FX Coudert <fxcoudert@gmail.com>, patch <gcc-patches@gcc.gnu.org>,
	ian@airs.com
Subject: Re: PING [4.1 regression, patch] build i686-pc-mingw32
Message-ID: <20050719201404.GA26817@trixie.casa.cgf.cx>
Mail-Followup-To: gcc@gcc.gnu.org, dannysmith@users.sourceforge.net,
	FX Coudert <fxcoudert@gmail.com>, patch <gcc-patches@gcc.gnu.org>,
	ian@airs.com
References: <42DCC2D5.5010900@gmail.com> <20050719195238.GB23513@trixie.casa.cgf.cx> <42DD5C82.4040502@gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <42DD5C82.4040502@gmail.com>
User-Agent: Mutt/1.5.8i
X-SW-Source: 2005-07/txt/msg00846.txt.bz2

On Tue, Jul 19, 2005 at 10:03:14PM +0200, FX Coudert wrote:
>>I'd prefer that Danny review this but neither of us has the right to
>>approve this patch.
>
>Well, then, who has the right to approve such a patch?
>
>>However, it seems like you're adding extra stuff to the Makefile where
>>it is already trying to do the right thing if $(LN) fails.  Couldn't LN
>>just be declared as "cp" for mingw or, alternatively, couldn't it be
>>defined as some failing command so that the cp in the failing condition
>>would be invoked?
>
>Well, perhaps I wasn't clear enough on what is the problem:
>
>-	case "$<" in \
>-	  ./*) ;; \
>-	  ../*) \
>-	     echo $(LN) $< as$(exeext); \
>-	     $(LN) $< as$(exeext) || cp $< as$(exeext) ;; \
>-	  *) echo '#!$(SHELL)' > as; echo 'exec $< "$$@"' >> as ; \
>-	     chmod +x as ;; \
>+	case "$(build)" in \
>+	  *mingw32*) \
>+	    cp $< as$(exeext) ;; \
>+	  *) \
>+	    case "$<" in \
>+	      ./*) ;; \
>+	      ../*) \
>+	        echo $(LN) $< as$(exeext); \
>+	        $(LN) $< as$(exeext) || cp $< as$(exeext) ;; \
>+	      *) echo '#!$(SHELL)' > as; echo 'exec $< "$$@"' >> as ; \
>+	        chmod +x as ;; \
>+	    esac \
>
>The cases with ./*) and ../*) already well handled, and just kept "as 
>is" when $(build) is mingw32. The $(LN) command succeeds, that's not the 
>problem.
>
>Only the remaining case, which creates a shell script, fails on mingw32. 
>The shell script is spawned to call the assembler, which fails. Only 
>real win32 executables can be spawned.

Ok.  Given that 'cp' was an acceptable fallback in the original version
of the above script, I wonder why 'cp' wasn't used instead of creating a
shell script wrapper.

cgf