buildworld Stop in /usr/src/usr.sbin/bsnmpd/bsnmpd

[Kevin Oberman]

> Date: Thu, 23 Mar 2006 09:06:33 +0100 (CET)
> From: Harti Brandt <hartmut.brandt at dlr.de>
> Sender: owner-freebsd-current at freebsd.org
> 
> On Wed, 22 Mar 2006, othermark wrote:
> 
> o>Kevin Oberman wrote:
> o>> > Date: Wed, 22 Mar 2006 14:24:08 -0800
> o>> > From: Jason Evans <jasone at FreeBSD.org>
> o>> > 
> o>> > Kevin Oberman wrote:
> o>> > 
> o>> > > > From: othermark <atkin901 at yahoo.com>
> o>> > > > Date: Wed, 22 Mar 2006 12:13:14 -0800
> o>> > > > Sender: owner-freebsd-current at freebsd.org
> o>> > > > 
> o>> > > > I'm scratching my head on this one, trying to buildworld with today
> o>> > > > sources
> o>> > > > to get the sendmail fix:
> o>> > > 
> o>> > > 
> o>> > > I am seeing the same thing. I read the tread on this from back on March
> o>> > > 7-9 and have confirmed that the "fix" is in my Makefile.inc1, but I
> o>> > > still can't make buildworld.
> o>> > > 
> o>> > > My last update was in the critical perion...Feb. 23, but I have been
> o>> > > unsuccessful in getting past it.
> o>> > > 
> o>> > > I have tried explicitly running the newly built getsnmptree (from
> o>> > > /usr/obj) and then building, but I have had no luck to this point.
> o>> > > 
> o>> > > Not only did the fix to Makefile.inc1 not do the trick, but
> o>> > > hand-building gnesnmptree and hand executing the commands to make the
> o>> > > oid.h file didn't help a bit.
> o>> > > 
> o>> > > I'm baffled. (And probably missing the obvious.)
> o>> > 
> o>> > This isn't a very satisfying solution, but if you do the buildworld with
> o>> > MALLOC_OPTIONS=jZ, you will avoid the gensnmptree bug.
> o>> > 
> o>> > Jason
> o>> > 
> o>> 
> o>> 
> o>> Jason,
> o>> 
> o>> Thanks! It worked like a charm.
> o>> 
> o>> Any explanation as to why zeroing a malloc makes this work (as opposed
> o>> to filling with 0xa5)? I'm sure confused.
> o>
> o>Double thanks, I was finally able to get a successful buildworld.   What is
> o>the likelyhood of other people hitting this?   This boxes' -current was only
> o>27 days old, before I tried this update.
> 
> It should not, because gensnmptree is now a buildtool. To find out what 
> goes wrong we need the buildworld log that Ruslan mentioned.

Unfortunately, by the time I saw Ruslan's message requesting the
buildworld output I had already built bsnmpd with the jZ malloc options
and now it just works. To get the failure again, I'd probably have to
roll my system back to Feb 23 again and then update and I just don't
have the time right now to do it. It's not a very fast system and
buildworld takes about 4 hours. It is also a remote system and I don't
have physical access to it until next Tuesday.

Sorry. Maybe someone else who stumbles over this can get a full
buildworld, but it probably only applies to people who built a system
from HEAD in the last week or so of February.
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634