From mboxrd@z Thu Jan  1 00:00:00 1970
Return-Path: <gdb-return-22963-listarch-gdb=sources.redhat.com@sourceware.org>
Received: (qmail 4423 invoked by alias); 2 Nov 2005 14:53:04 -0000
Mailing-List: contact gdb-help@sourceware.org; run by ezmlm
Precedence: bulk
List-Subscribe: <mailto:gdb-subscribe@sourceware.org>
List-Archive: <http://sourceware.org/ml/gdb/>
List-Post: <mailto:gdb@sourceware.org>
List-Help: <mailto:gdb-help@sourceware.org>, <http://sourceware.org/ml/#faqs>
Sender: gdb-owner@sourceware.org
Received: (qmail 4394 invoked by uid 22791); 2 Nov 2005 14:53:01 -0000
Received: from nevyn.them.org (HELO nevyn.them.org) (66.93.172.17)
    by sourceware.org (qpsmtpd/0.30-dev) with ESMTP; Wed, 02 Nov 2005 14:53:01 +0000
Received: from drow by nevyn.them.org with local (Exim 4.54)
	id 1EXJz1-0007P0-0S; Wed, 02 Nov 2005 09:52:59 -0500
Date: Wed, 02 Nov 2005 14:53:00 -0000
From: Daniel Jacobowitz <drow@false.org>
To: Wu Zhou <woodzltc@cn.ibm.com>
Cc: gdb@sources.redhat.com
Subject: Re: The root cause for SEGV in evaluating fortran function call, any solution or suggestion?
Message-ID: <20051102145258.GA28372@nevyn.them.org>
Mail-Followup-To: Wu Zhou <woodzltc@cn.ibm.com>, gdb@sources.redhat.com
References: <Pine.LNX.4.63.0508221743470.10411@linux.site> <Pine.LNX.4.63.0511021033530.3549@linux.site>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Pine.LNX.4.63.0511021033530.3549@linux.site>
User-Agent: Mutt/1.5.8i
X-SW-Source: 2005-11/txt/msg00033.txt.bz2

On Wed, Nov 02, 2005 at 10:43:03AM +0800, Wu Zhou wrote:
> Hi all,
> 
> I had found the reason why gdb will drop into SEGV when evaluating the 
> fortran function calls.  In g77 (gfortran might be the same), when we try 
> to call FUNC_NAME (ARGS), the ARGS is passed as the pointer to the real 
> parameters.  
> 
> While we issue "print FUNC_NAME (ARGS), these ARGS are passed as the 
> original types.  So fortran code can't handle that and SEGV occurs.  
> Considering this special argument-passing mechanism, do we have any 
> workaround for it in gdb's evaluation code?  Create a dummy memory address 
> for the arguments and pass that address instead?  or any others? 

I don't think there's any way to handle that right now: you'll have to
add one.  I am not sure where in the call-function process it should
be.  Probably, you will also need a way to allocate them - on the stack
would be better than malloc, which is how gdb usually allocates memory.

-- 
Daniel Jacobowitz
CodeSourcery, LLC