X-Spam-Check-By: sourceware.org
From: "Dave Korn" <dave.korn@artimi.com>
To: <cygwin@cygwin.com>
Subject: RE: Fortran calling c compilation problem
Date: Wed, 2 Aug 2006 15:11:35 +0100
Message-ID: <050501c6b63d$8ffc6160$a501a8c0@CAM.ARTIMI.COM>
MIME-Version: 1.0
Content-Type: text/plain; 	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Mailer: Microsoft Office Outlook 11
In-Reply-To: <44CF677A.3050707@statslab.cam.ac.uk>
Mailing-List: contact cygwin-help@cygwin.com; run by ezmlm
Precedence: bulk
List-Unsubscribe: <mailto:cygwin-unsubscribe-archive-cygwin=delorie.com@cygwin.com>
List-Subscribe: <mailto:cygwin-subscribe@cygwin.com>
List-Archive: <http://sourceware.org/ml/cygwin/>
List-Post: <mailto:cygwin@cygwin.com>
List-Help: <mailto:cygwin-help@cygwin.com>, <http://sourceware.org/ml/#faqs>
Sender: cygwin-owner@cygwin.com
Mail-Followup-To: cygwin@cygwin.com
Delivered-To: mailing list cygwin@cygwin.com

On 01 August 2006 15:39, Ben Wright wrote:


> g++ -mno-cygwin -Wl,--add-stdcall-alias
> -I/cygdrive/c/Java/jdk1.5.0_07/include
> -I/cygdrive/c/Java/jdk1.5.0_07/include/win32 -shared -o Samplers.dll <a
> whole bunch of source files here> -lstdc++ -lg2c
> 
> /cygdrive/c/DOCUME~1/ben/LOCALS~1/Temp/ccWWbiIF.o:gengam.f(.text+0x5):
> undefined reference to `_dgamdv_'
> (plus others of similar form)
> 
> The linker can't seem to find references of the form _dgamdv_. Functions
> are defined in the c portion as gdamdv_(<whatever>), and are called from
> Fortran as DGAMDV(). This works fine under Linux.

  Not quite sure why you would expect a function called _dgamdv_ to match one
which is called gdamdv_XXXXXX.  They appear to have very different names, and
one has a suffix where the other doesn't.  Is there some kind of typo in this
para?
 
> What's baffling is that if I compile just the c portion and inspect the
> output with nm then the supposedly undefined references are in there.
> 
> 10004148 T _dgamdv_

  However, this suggests that the real problem is just that the source files
are in the wrong (meaning non-dependency) order on the compiler command line,
and hence on the linker command line.

> Is there some quirk of cygwin when calling c from within fortran I need
> to set a flag for when compiling? I'd ordinarily chalk it up to my weak
> knowledge of c- but it compiles without complaint under Linux.

  This works fine on Linux where a shared lib can contain undefined references
that are linked up at runtime, but it doesn't work on 'doze where a dll has to
be fully resolved.  To fix it, you could just reorder the source files into
dependency order on the command-line, so that any file that refers to an
undefined symbol comes *before* the file that supplies that symbol, or you
could separate compiling and linking into two separate stages, and use the
--start-group and --end-group options to wrap all the object files on the
linker command line.

    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....


--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/