Bug 44016
| Summary: | 4.7R installworld fails with /usr/obj readonly | ||
|---|---|---|---|
| Product: | Base System | Reporter: | Jed Clear <clear> |
| Component: | misc | Assignee: | ru <ru> |
| Status: | Closed FIXED | ||
| Severity: | Affects Only Me | ||
| Priority: | Normal | ||
| Version: | Unspecified | ||
| Hardware: | Any | ||
| OS: | Any | ||
|
Description
Jed Clear
2002-10-13 18:30:00 UTC
On Sun, Oct 13, 2002 at 10:23:14AM -0700, Jed Clear wrote:
> >Description:
> If /usr/obj is mounted read only, make installworld fails during Perl.
>
> Normally I update my firewall by NFS mounting /usr/src and /usr/obj read-only, off a server with a lot more compile horsepower. [Ever do a buildworld on a 486/66?] This time, while doing the installworld to upgrade from 4.5p19 to 4.7R, it failed during perl5.
>
> [The following are from notes that I can hardly read so may contain mistakes and the ??? is something I can't read. :-(]
> It failed with:
> Appending installation info to /usr/libdat/perl/5.00503/mach/perl/???pod===>gnu/usr.bin/perl/library/SBM_FILE
> Stop in /usr/obj/usr/src/gnu/usr.bin/perl/library/SDBM_File/ext/SDBM_File/sdbm
>
> Mounting /usr/obj r/w lets the installworld proceed, but seems wrong to me.
>
> This was a freshly SUP'd 4.7R, as of 10/12.
> >How-To-Repeat:
> Mount /usr/obj readonly
> cd /usr/src; make installworld
> >Fix:
> WORKAROUND is to mount /usr/obj read/write.
>
> It looks like the fix will involve how Perl is installed.
Get your system clock in sync! (Or just run adjkerntz -i after boot if your
CMOS clock are set to localtime).
Best Regards,
Serg N. Voronkov,
Sibitex JSC.
State Changed From-To: open->closed The usual src/ modtimes / system time mismatch. The two clocks can't be any more in sync. I run ntpd on both. The firewall acts as NTP server to the fileserver. The firewall is synched to an Internet stratum 2 source (not that that matters as we're talking relative time between the two boxes). Both have /etc/wall_cmos_clock and run adjkerntz -i in /etc/rc. The timezones on both remain at EDT and the firewall keeps the right timezone when shutdown to single user. Now theoretically the clocks could drift after the firewall had been shutdown to single user for the installworld, but that would be a hell of a drift rate. I suppose if the 4.5 adjkerntz couldn't work with the new 4.7 kernel, that could be the problem. But then I'd expect a lot more to try to build, or it to fail earlier. If I'm missing something with time, let me know, otherwise please reopen the bug. -Jed State Changed From-To: closed->open Not a clock problem. State Changed From-To: open->closed I've just checked it here, the upgrade went fine: # uname -a FreeBSD gatekeeper.sunbay.crimea.ua 4.6-RC FreeBSD 4.6-RC #0: Wed May 29 14:02:28 EEST 2002 root@perl.sunbay.crimea.ua:/usr/obj/usr/src/sys/GATEKEEPER i386 # mount /dev/da0s1a on / (ufs, local) /dev/da0s1f on /usr (ufs, local) /dev/da0s1e on /var (ufs, local) procfs on /proc (procfs, local) perl:/usr/src on /usr/src (nfs, read-only) perl:/usr/obj on /usr/obj (nfs, read-only) # make installworld installkernel # reboot # uname -a FreeBSD gatekeeper.sunbay.crimea.ua 4.7-STABLE FreeBSD 4.7-STABLE #0: Tue Oct 15 12:36:24 EEST 2002 ru@perl.sunbay.crimea.ua:/usr/obj/usr/src/sys/GATEKEEPER i386 Other possible reasons for a failure include: some of the sources have modification time set to the future (use find(1)); make.conf options on the building and installing hosts do not match. See what ``make -n'' is trying to build in the offending directory. Responsible Changed From-To: freebsd-bugs->ru |
