Back to the Future - 64-bit time_t on sparc64

[Garance A Drosihn]

At 1:31 PM -0600 2/16/04, Tillman Hodgson wrote:
>  > Garance A Drosihn wrote:
>>  >
>>  >Well, I have done more typing and testing, and it looks like I
>>  >can reliably upgrade a 32-bit time_t system to 64-bit time_t
>>  >even for people who install from NFS-mounted partitions.
>>  >
>>  >I have also expanded the instructions for updating, and tried
>>  >to make them more useful and informative.  So there is now the
>>  >writeup, and two useful scripts:
>>  >
>>  >http://people.freebsd.org/~gad/time-64/UPDATING.64BTT
>>  >http://people.freebsd.org/~gad/time-64/installworld_oldk
>>  >http://people.freebsd.org/~gad/time-64/installworld_newk
>>  >
>>  >The 'oldk' script is only needed for installing via NFS mounts.
>  > >The 'newk' script is recommended for anyone doing this upgrade.


>When you say "installing via NFS mounts", are you referring
>to /usr/obj?

The extra instructions will be needed if there is *any* partition
that you need to have NFS-mounted at the time you enter the
'make installworld' command.

Usually when you are upgrading, you can reboot into single-user
mode on the "new kernel" (after doing the 'make installkernel'),
and then simply NFS-mount all the directories that you need for
the installworld step.  However, when moving from 32-bTT to
64-bTT, you will find it impossible to NFS-mount anything after
that reboot, unless you do the extra ./installworld_oldk step
before that reboot.

>The reason that I ask that my /usr/src is remote but my /usr/obj
>is local.  If /usr/src being remote marks me as doing an NFS
>install for the purposes of these scripts, what are the
>disadvantages (if any) of use the _oldk script?

Yes, your example is an "NFS install", wrt this change.

For this specific update (going from 32-bTT to 64-bTT), there is
just a slight increase in risk by running the installworld_oldk
script.  If you follow the directions, and if you are ONLY making
the update from 32-bTT to 64-bTT (as described in the directions),
then the risk is pretty close to zero.  If you think you're going
to do the 64-bTT change, and AT THE SAME TIME also 'cvsup' another
three weeks-worth of other changes to 5.2-current, then the risk
could be much more serious.

This is described in a little detail in the UPDATING.64BTT file.
Actually there are two parts to the installworld_oldk script.
The first part (creating /boot/kernel/bin) is totally safe.  It's
the second part which introduces a little risk.

>  > >Adventurous people are invited to try this for installs on any
>  > >system, and let me know how it goes
>
>I plan up upgrade caliban.rospa.ca in the near future and I'll
>report back how it goes.

Thank you.  The more people who can test this, the better.

(aside:  "64bTT" is my abbreviation for "64-bit time_t")

-- 
Garance Alistair Drosehn            =   gad at gilead.netel.rpi.edu
Senior Systems Programmer           or  gad at freebsd.org
Rensselaer Polytechnic Institute    or  drosih at rpi.edu