Going from 8.0-CURRENT to RELENG_7

[eculp at encontacto.net]

Quoting "Alexandre \"Sunny\" Kovalenko" <alex.kovalenko at verizon.net>:

> This is basically a warning and an advise -- as usual -- YMMV.
>
> I have missed HEADS-UP that RELENG_7 has been branched and found myself
> running 8.0 after cvsup/buildworld/installworld cycle. Since I wanted to
> stick to RELENG_7 and since branching occured 2 days ago, I've figured I
> should be able to change tag in SUPFILE cvsup and rebuild.
>
> Hence the warning: *don't do that* -- buildworld failed midway with
>
> /libexec/ld-elf.so.1: Undefined symbol "__mb_sb_limit" referenced from
> COPY relocation in sh.
>
> and you are left with non-working system.
>
> Now the advice part: *do not despair* -- you can boot into single user
> mode, use /rescue/sh, manually copy
>
> /bin/sh
> /usr/bin/make
> /usr/bin/awk
> /usr/sbin/mtree
> /usr/sbin/zic
> /lib/libedit.so.6
> /lib/libncurses.so.7
>
> from your OBJ directory, then do 'make installworld' twice (something
> non-important failed in the first pass) and you should be back in
> business.
>
> Hopefully, everyone on the list reads it diligently and does not need
> such advice, but if you shot yourself in the foot, here is the chance to
> avoid reinstalling the system from scratch.

Here is another one who didn't :( I realized it after seeing your  
email.  Here is what I did which now I see is what you did.
   1. Changed the configuration of my cvs-supfile to:

      *default tag=RELENG_7
      *default base=/usr/local/etc/cvsup
      *default prefix=/usr
      *default release=cvs
      *default delete use-rel-suffix
      src-all
      doc-all
      ports-all
      cvsroot-all

I had built and installed but not rebooted Current 8.0 for two or  
three previous days.  It successfully built and installed world and  
kernel.  it initiated installworld and broke while installing man.   
This happend to me on two machines one local and one remote location.   
The local machine is back up but I made a mess of it.  The rescue  
suggestion didn't work because it was the same as the installed  
programs.  I tried booting with a pre-8.0 kernel and same problem so I  
installed a cd that I had form the 5th of October and lost etc and a  
few other files but managed to  get it back together. I have erased  
src and am re cvsuping with the above cvs-supfile and plan to follow  
the build world, build kernel, install kernel, install world, but I am  
concerned that that may not be the best solution.  I certainly don't  
want to do this again. ;)

Now for the machines that are in the same situation but located far away.

On is in the same situation as this and is the closest so I can go  
there tomorrow and do the same procedure unless there is something  
that I can do  from here.

The other three machines are plane rides.  Luckily, I stopped the  
RELENG_7 build before it broke but the did all finish the sup.  I'm  
thinking that the smart thing may be to cvsup again but with current  
8.0 tag and just build current and reboot and wait to see if I can get  
back and if not just ride out current.  That sure beats spending the a  
lot of time and money traveling.

Any and all suggestions appreciated for both cases and if you have any  
inclinations/observations on the stability of 8.0 if I rebuild and  
reboot the machines.

Thanks for your help,

ed