From owner-freebsd-stable Sat Dec 21 1:22:23 2002 Delivered-To: freebsd-stable@freebsd.org Received: from mx1.FreeBSD.org (mx1.freebsd.org [216.136.204.125]) by hub.freebsd.org (Postfix) with ESMTP id DEB6F37B401 for ; Sat, 21 Dec 2002 01:22:20 -0800 (PST) Received: from h33.erkki.ton.tut.fi (h33.erkki.ton.tut.fi [193.166.84.86]) by mx1.FreeBSD.org (Postfix) with ESMTP id C699443EDA for ; Sat, 21 Dec 2002 01:22:09 -0800 (PST) (envelope-from chu@gpi.ru) Received: from gpi.ru (chu@localhost [127.0.0.1]) by h33.erkki.ton.tut.fi (8.12.3/8.12.3) with ESMTP id gBL9L4CF007035 for ; Sat, 21 Dec 2002 11:21:10 +0200 (EET) (envelope-from chu@gpi.ru) Message-ID: <3E04327C.D7948A41@gpi.ru> Date: Sat, 21 Dec 2002 11:21:01 +0200 From: Vladimir Chukharev X-Mailer: Mozilla 4.79 [en] (X11; U; Linux 2.4.2 i386) X-Accept-Language: ru, en, fi MIME-Version: 1.0 To: freebsd-stable@freebsd.org Subject: Re: The plot thickens (problem solved!) (was Re: More information ...) Content-Type: text/plain; charset=koi8-r Content-Transfer-Encoding: 7bit Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG I can share my experience with this problem. I got the same kind of linking errors for most ports with libXt (IIRC, it was libXt, not just libX) after upgrading XFree86* via portupgrade with -O3 in my make.conf. I had problems with X ports not first time, I remember that it is clearly said do not report untill you reproduce it without optimization. So I tried to remove optimization and started to rebuild everything without -O3. Two previous times I removed /usr/X11R6 and successfully reinstalled X and then all needed packages. When I reported success in followup to PR, the PR was closed. (Well, I said it can be closed, if there is no interest in hunting this bug, which hunting is hardly a fun.) This time I managed to coupe by forced reinstall of x11/XFree86-4-libraries, x11-servers/XFree86-Server and x11/XFree86-4-clients in a particular order. The forced reinstall of XFree86 metaport does not help. I don't remember now the correct order, but first I tried two other orders without success. I believe there is a bug in mutual dependencies of these ports. It is exposed only when updating to incompatible version. In this case incompatibility was induced by -O3. I don't know if this can be of any help... Best regards, V.Chukharev > > > Later attempts _without_ optimalisations resulted in the same errors. I > > > did use portupgrade -fR, and I _do_ keep a clean and tidy system. I > > > cvsup religiously, rebuild world/kernel when needed, and read UPDATING > > > etc. etc. There's absolutely no reason why this couldn't have affected > > > someone else as well. Even from a clean environment (everything X related > > > removed, as in pkg_delete all of X and nuking what's left as in rm -rf > > > /usr/X11R6) this glitch occured *again*. Odd, no? > > > > > > Oh, I must've recompiled the whole (and parts) of X like 10 times today. I > > > also rebuild world and kernel again just to make sure my toolchain and > > > evironment was pristine. > > > > OK, that's good to know. I didn't see these facts before in your other mails. > > I might have taken my practice of keeping my system as tidy as possible > for granted with other setups. I just assumed everyone's system was as > tidy as mine :) > > [snip] > > > > I might be the first to have noticed. What if suddenly 20 more people in > > > the span of the next 24 hours report that X or X apps won't build? > > > > That's of course true, but 16 hours have passed since your first email > > and noone else reports such a problem. Experience has shown that X build > > problems will probably be reported more frequently here or on -ports > > during that period if they affect all of the FreeBSD user group. > > Or maybe everyone's on holiday of course, I don't know... > > Hey, it's not every day one gets to find a bug in Xfree/FreeBSD. I jumped > at it. Just doing my bit. Maybe I did overdo it a bit... > > > What I'm trying to say, politely, is something like 'patience, young jedi', > > I think you will get more results from the lists that way. No offense > > meant. > > None taken. > > > BTW, I also didn't see a response to Joe Marcus Clarke's questions: > > > > > What ports are broken because of this? Any port which can't find > > > -pthread or -lc_r on its own should have ${PTHREAD_LIBS} added to its > > > LIBS configure argument. With that, there are also some thread-related > > > CFLAGS that should be added. These are defined by the ports system as > > > ${PTHREAD_CFLAGS}. > > > > Except for your test program, what ports don't work? What lead you to > > the initial conclusion that something is wrong with X? > > All ports that wanted to link with -lX11 broke here. I suspected that X > was the problem when I digged through the autoconf config.log files > which stated everytime that libXThrStubs.so.6 had unresolved symbols, so I > started grepping for those symbols in the X lib source. > > When I found them, I found a UIThrStubs.c.orig file which meant that > that file was patched. I put two and two together and looked at what was > patched, and concluded that that was the source for the unresolved > symbol problems. When I undid the patch to that file, all was well and > linked happily again. Even my test code worked as expected. > > So there you have it :) > > Cheers, > Emiel To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message