Newer installworld_nk script (64-bit time_t)

Garance A Drosihn drosih at
Sat Jan 24 02:02:00 PST 2004

At 10:42 PM -0500 1/21/04, Garance A Drosihn wrote:
>At 3:41 AM -0500 1/12/04, I (Garance) wrote:
>>Is the magic script.  Probably a little more elaborate than it
>>needs to be, but I had plenty of extra time while waiting for
>>'make buildworld' to finish...
>Fwiw, I just made to minor changes to this script, to make it a
>little more user-friendly.  The option for "yes, go ahead and
>do the installworld" actually works now...

Well, thanks to a combination of events, I ended up having to a
few extra hours to work on this, so I've gone through another
few rounds of improvements to the script.

Is the newest version of the script.  It has not been used
on an "upgrade sparc64 from 32BTT to 64BTT" run, but it has
been extensively tested with a few dozen installworlds on
my dual-athlon system (which is just a *little* faster than
my Ultra-10).

By being more and more restrictive with my PATH settings, I
tracked down a few more files that are prudent to copy, and I
realized that some steps should be done in a different order.
I also added a -S option, which causes the script to use
symlinks to most executables it is pulling from /usr/obj,
instead of copying them all.  So, even though I'm bringing in
more files, the /tmp/install-nk* directory is now only 200K to
250K when -S is used.  I also added more tests and improved
some of the script-debugging features.  There is also a -M
option, which causes the installworld to be done with PATH
set to *only* the /tmp/install-nk* directory.

I also tested this version (on i386) by:
     boot into single user
     script /var/tmp/somefile
     make installworld         <- standard make target
     script /var/tmp/otherfile
     ./installworld_nk         <- magic alternative
repeating that several times with different options to the script.
I then did 'diff's on all the outputs, to make sure that they all
seemed to do exactly the same thing.  As far as those script
outputs are concerned, the same commands were executed with the
exact same outputs.

Older versions of the script are still available in that same
directory, but right now this is the most heavily-tested one
(except that it hasn't actually been used for a 32->64 upgrade).

I think it's an interesting exercise to try to make this script
as bullet-proof as possible.  I'm sure that it's adequate for
immediate need, but I'd like to see how far it could go.

Garance Alistair Drosehn            =   gad at
Senior Systems Programmer           or  gad at
Rensselaer Polytechnic Institute    or  drosih at

More information about the freebsd-sparc64 mailing list