Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 5 Jun 1996 22:21:36 -0600
From:      Nate Williams <nate@sri.MT.net>
To:        stable@FreeBSD.org, committers@FreeBSD.org
Subject:   Status of -stable
Message-ID:  <199606060421.WAA24403@rocky.sri.MT.net>

next in thread | raw e-mail | index | archive | help
I thought Jordan was going to post something, but I suspect he passed
out from stress and lack of sleep, so I'll attempt to recap what the
status is, and give a summary of what happened.

I *think* everything should be working again (including src/eBones and
src/secure) as it was before the mega-commit episode, but you may need
to wait about 24 hours for the changes to get propogated to you via the
different distribution channels.

In order to get things working correctly you *must* do a make world if
you do any significant builds in the last 2 weeks.  With the
library/make/etc.. changes that are now backed out, the tree is
significantly different from what it was 36 hours ago.

Note, *IF* you ended up with libc.so.3.0 built on your system, you are
in a a bit of a sticky situation.  There is no longer version 3.0 of the
library, as it's been reverted back to the 2.2 version.  However, *IF*
you have this shlib on your system, you can't delete it because all of
your binaries have been linked against it, and if you don't delete it
none of the binaries will link against the older version.

So, my advice to you is to go into single user-mode, delete the 3.0
libary, and then try to bootstrap yourself with a 'make world', and
restore any binaries that give error messages with the versions from
2.1R.

Or, you could take the easy route and re-install the binaries from 2.1R
and then do a make world, which should bring you up to -stable in a
hurry.

Sorry about the breakage, and normally this sort of thing wouldn't have
happened in stable but it's getting close to release time and the
release engineer wanted to get everything in that he could.  As with
most projects, it turned out to be *much* more difficult than originally
anticipated, so rather than forge ahead with a tree that was slightly
broken for a couple weeks it was decided to back out everything.  The
hard part comes in that changes were being made during the time the
userland merge started and the time it was decided it was too much
hassle.  So, backing things up wasn't too easy, and unfortunately CVS
didn't make it any easier due to some mis-features that *should* be
fixed.

As for what happened with the secure bits, it turns out that CTM doesn't
send out any of the 'non-exportable' bits (which happens to be
src/eBones and src/secure), so when backing out the changes using code
restored via CTM it ended up wiping out the bits from stable.  Luckily,
John Polstra had a backup copy of the CVS bits from May 21 which was
restored, and there was only one commit made since that time which was
added back into the tree, so things should be *exactly* the same shape
as it was before the mega-commits.

Other than that, I backed out a couple of commits that happened in the
kernel sources that reverted out some kernel changes accidentally, but I
*think* those were the only bad changes.  I reviewed almost all of the
include file changes, and some of the library changes and they appear to
be fine, but if any of you see some breakage let us know, or if you
think something is the wrong version please pipe up.

[ But please wait until *after* the bits have settled down, or you are
*SURE* that you have the most current bits from this afternoon.  ]

Anyway, I think -stable is now stable once again.



Nate



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199606060421.WAA24403>