Chris Lumens wrote:
Use zonetab module for in place of zone.tab file.
Zonetab (used by UI and s-c-date) adds some of Etc/* specifiers.
This looks fine to me, and fixes up a lot of bugs I think. Can we
consider this one for backporting to RHEL5 if those bugs start to become
really important?
Here is the timezone setting situation:
After my patches for timezone checking
(https://bugzilla.redhat.com/show_bug.cgi?id=404321), where I use
zone.tab in Fedora, and /usr/share/zoneinfo tree in RHEL, there was
regression bug filed about not allowing specifiers allowed before.
My checking was meant to be done also to have valid specifier for setting
correct systemtime in setuptime step (needed e.g. to have correct - esp.
not
in future - timestamps of files). And setuptime step needs zoneinfo files.
Without the check, we failed to set systemtime, but if the specifier was
present in tzdata package, timezone of installed system was configured
ok (if
not, /etc/localtime was just missing in the system), and so many specifiers
(like /Etc/...) could be used in ks to install succesfully.
The bug filed in RHEL is:
https://bugzilla.redhat.com/show_bug.cgi?id=481617 which is really a
regression, perhaps the checking shouldn't have been used in rhel, I should
have only get rid of traceback and fail to set systemtime silently during
install.
There is also Fedora clone of the bug
https://bugzilla.redhat.com/show_bug.cgi?id=483094,
but in Fedora we limit allowed specifiers in documentation.
In RHEL I'd rather fix the bug by leaving whole zoneinfo tree in stage2,
not only
files specified by TIMEZONES in upd-instroot (means growing stage2 by
about 200-300KB).
Dealing with the checking, i'd like to fix also systemtime setting thing
14:01:41 INFO : moving (1) to step setuptime
14:01:41 ERROR : unable to set timezone
- we do it without having needed timezone files (Fedora):
https://bugzilla.redhat.com/show_bug.cgi?id=461526
- we do it too late (both in Fedora and Rhel):
https://bugzilla.redhat.com/show_bug.cgi?id=6175
And I'd like to do it by:
- including /usr/share/timezone tree (or at least subtree containing
specifiers in /zone.tab) back into stage2 (about 650 KB). The only
other ways would be to move setuptime step after packages
install and using installed files (setting systemtime at this
stage is late, but can be useful at least for %post scripts); or to
have/generate something like zone.tab containing info for TZ value
setting needed for hwclock --hctosys call.
- moving setuptime step after accounts step
I tried it and it solved the bugs for me.
I know we have been moving away from having complete /usr/share/zoneinfo
in stage2, but I think it is the simplest=>best way to fix setuptime issues
and it would fix also ks value checking issue.
Would you be ok with having /usr/share/zoneinfo tree in stage2?
Radek
_______________________________________________
Anaconda-devel-list mailing list
Anaconda-devel-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/anaconda-devel-list
