From owner-freebsd-hackers Mon Mar 19 17:12:35 2001 Delivered-To: freebsd-hackers@freebsd.org Received: from mx7.port.ru (mx7.port.ru [194.67.23.44]) by hub.freebsd.org (Postfix) with ESMTP id 1BF5C37B73F; Mon, 19 Mar 2001 17:12:27 -0800 (PST) (envelope-from kabaev@mail.ru) Received: from adsl-141-154-120-34.bostma.adsl.bellatlantic.net ([141.154.120.34] helo=kan.dnsalias.net) by mx7.port.ru with esmtp (Exim 3.14 #15) id 14fAhB-0005KQ-00; Tue, 20 Mar 2001 04:12:22 +0300 Received: (from kan@localhost) by kan.dnsalias.net (8.11.3/8.11.3) id f2K1CJM09332; Mon, 19 Mar 2001 20:12:19 -0500 (EST) (envelope-from kan) Message-ID: X-Mailer: XFMail 1.4.7p2 on FreeBSD X-Priority: 3 (Normal) Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 8bit MIME-Version: 1.0 In-Reply-To: <3AB697F1.4BF23D6E@pleach-hamburg.de> Date: Mon, 19 Mar 2001 20:12:18 -0500 (EST) From: "Alexander N. Kabaev" To: Titus von Boxberg Subject: Re: GCC Upgrade? Cc: hackers@FreeBSD.ORG, obrien@FreeBSD.ORG Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On 19-Mar-2001 Titus von Boxberg wrote: > David O'Brien wrote: >> >> On Mon, Mar 19, 2001 at 02:54:52PM +0100, Titus von Boxberg wrote: >> > Since at least aug. 2000 (according to the mailing list >> > archives) the exception handling in base system g++ is broken >> > (at least for multithreaded programs) The problem you are talking about has nothing to do with threads. Rather, the problem was with the way in which GCC handles inline function expansions. > > Sorry, I don't have yet isolated the problem. > But this link seems to describe exactly the same problem. > "http://www.FreeBSD.org/cgi/getmsg.cgi?fetch=298763+300854+ > /usr/local/www/db/text/2000/freebsd-hackers/20000716.freebsd-hackers" I thought GCC 2.95.3 has this problem fixed, but it turns out I was wrong. I've had patches to fix this particular breakage for ages now and I even offered them to David O'Brien on more than one occasion. The fix has been posted on the gcc-devel mailing list and Berndt Schmidt even included it into some of GCC 2.95.3-testXX release. And that time I decided that my job is done, but apparently Berndt managed to revert (most likely, by mistake) before release. I think my relative unrelated changes have been killed as part of the bigger sjlj exceptions rewrite by Mike Henderson :( > you can search for "DWARF AND exception"; then you'll find the links. > > I'm using omniORB, and on their home page you can find a > hint in the docs describing the same thing (but with g++ on a AIX/RS6000). > Yet another problem I have patches for. GCC does not handle PIC register correctly when handling exceptions thrown across shared library boundaries. Additionally, the shared libraries in AIX make the challenge of sharing exception context state among all shared libraries in the executable very interesting. I was unable to get these fixes into official CVS source - FSF guys apparently decided that the problem cannot be fixed. They even seriously tried to convince me that my patch could not work except by coincidence without even taking a sigle look at the patch itself - even though that 'coincidence' reliably works for me in multithreaded CORBA server written using omniORB :) I basically gave up on them. Drop me a line if you need help getting GCC 2.95.3 work properly on AIX 4.x. > Before end of april I cannot investigate the problem any further. > please let me know by then if I may help you with that problem. > > regards > titus > > To Unsubscribe: send mail to majordomo@FreeBSD.org > with "unsubscribe freebsd-hackers" in the body of the message ---------------------------------- E-Mail: Alexander N. Kabaev Date: 19-Mar-2001 Time: 19:05:53 ---------------------------------- To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message