Mailing-List: contact cygwin-help@sourceware.cygnus.com; run by ezmlm
Sender: cygwin-owner@sources.redhat.com
Message-ID: <001501c057c5$ea36fd20$9865fea9@intel9txl52pzr>
From: "Tim Prince" <tprince@computer.org>
To: <cygwin@sources.redhat.com>
Subject: g++/binutils fragility
Date: Sun, 26 Nov 2000 08:28:29 -0800
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit

Sorry to be posting a problem without suggesting a solution.  I have a piece of C++ code which I have received from one of my
co-workers.  Making a few changes to make it acceptable to g++, without changing its behavior on the original compilers, I find that
it produces a SIGSEGV  under the stock cygwin installation (repeatedly so, over a number of setup.exe refreshes).  I grabbed the
binutils-20001029-2 source and built and installed it; the problem went away.  I set the alignment parameter in bfd/coff-i386.c to
(3) (64-bit alignments) and it re-appeared. The thing runs reliably in RH linux with gcc-2.95.2, but not with the commercial alpha
test compiler which I have.

I have been building binutils with CFLAGS='-Os -march=pentiumpro -pipe';  I've wondered whether there are any alignment problems
which would require more than one package to be built with identical compiler version and options and bfd alignment settings.

When I run under gdb, I am unsuccessful in attempting to step into the functions where the fault occurs; the cursor winds up on the
final "}" of one function or another.  This did not change with the latest experimental gdb.

I apologize for these struggles of an old Fortran programmer, but I have eliminated the warnings and complaints produced by the
various compilers, and still I have this effect: in summary,

 installing binutils from cygwin source produces reliable operation; installing the cygwin binary does not.

OS:  W2KSP1
IBM T20 256MB RAM


--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe@sourceware.cygnus.com