From owner-freebsd-current  Sat Jan 20 19:30:06 1996
Return-Path: owner-current
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id TAA10524
          for current-outgoing; Sat, 20 Jan 1996 19:30:06 -0800 (PST)
Received: from austin.polstra.com (austin.polstra.com [206.213.73.10])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id TAA10481
          Sat, 20 Jan 1996 19:29:55 -0800 (PST)
Received: from austin.polstra.com (jdp@localhost) by austin.polstra.com (8.6.12/8.6.12) with ESMTP id TAA15704; Sat, 20 Jan 1996 19:29:37 -0800
Message-Id: <199601210329.TAA15704@austin.polstra.com>
To: Nate Williams <nate@sri.MT.net>
cc: Douglas Thomas Crosher <dtc@scrooge.ee.swin.oz.au>, sos@FreeBSD.org,
        dima@best.net (Dima Ruban), freebsd-current@FreeBSD.org
Subject: Re: problems in -current (was Re: awk broken ???) 
In-reply-to: Your message of "Sat, 20 Jan 1996 19:39:13 MST."
             <199601210239.TAA18929@rocky.sri.MT.net> 
Date: Sat, 20 Jan 1996 19:29:36 -0800
From: John Polstra <jdp@polstra.com>
Sender: owner-current@FreeBSD.org
Precedence: bulk

Nate writes:
> Douglas Thomas Crosher writes:
> > > After updating to the lastest I see awk failing on make depend in
> > > sys/compile/XXXX:
> > > 
> > > sh ../../kern/vnode_if.sh ../../kern/vnode_if.src
> > > awk: cmd. line:4: (FILENAME=- FNR=318) fatal error: internal error
> > > abort - core dumped
> > > 
> > > Also my freshly built gcc-2.7.2 coredumps wih sig 11...
> > > 
> > > It works on a kernel from jan 6...
> > > 
> > > Anybody else having problems with the latest changes....
> 
> I have one idea what it may be.  It *appears* to be a library problem
> given that older libraries work.  Assuming it's a library problem, I can
> only see two significant library changes which have occurred recently.
> 
> 1) Peter's import of the new bind code.  This shouldn't affect gcc or
>    awk, so I doubt this.
> 
> 2) John Polstra's change which affected all shlibs.
> jdp         96/01/16 16:03:10
> 
>   Modified:    share/mk  bsd.lib.mk
>   Log:
>   Always link /usr/lib/c++rt0.o into a shared library.  CPLUSPLUSLIB is no
>   longer necessary, and can be removed from Makefiles.
> 
> Unfortunately, I suspect the latter, but have no way of checking it
> right now.

I have to admit that I thought of the same thing when these reports
started showing up in -current.  But it is really hard to see how
the change I made could account for these symptoms.  My change
causes a small piece of code (which does nothing) to be executed
when a shared library (such as libc) is first mapped into memory.
It seems like, if there was something wrong with these changes,
the symptom would appear immediately when you executed anything -- not
at some later time, and not with an "internal error" message.

I'm not saying it's impossible (been around too long for that).  But it
seems very unlikely to me.

-- John