>On Tue, Feb 05, 2008 at 09:12:55AM +0100, Casper.Dik at Sun.COM wrote:
>> I note that this is a lot worse for people who update biweekly,,
>> but I also wonder why postrun needs to run at standard priority?
>>
>> Some systems will run certain services faster and appear up more quickly
>> if postrun is run at a lower CPU priority.
>
>Would it be useful to be able to associate postrun with services?
>Think "update method". After an alt-root/LU upgrade every affected
>service would have it's update method run prior to its start method.
That would have been my next suggestion if you hadn't beaten me to it.
What seems natural to me is that each "postrun" script has a natural
dependency, e.g., GNOME or etc.
(fc-cache -> cde-login)
(cde-printinfo -> cde-login)
And that these should somehow be grouped; this would additionally allow
us to do partial initializations. No dtlogin, no fc-cache).
We also have a few services which do this too (cde-print-info, fc-cache,
ppd-cache-update) and I am assuming that these will all be moved
to the new scheme).
Are the "postrun" methods required to be idempotent? "SMC" carried
its own "postrun" method and ran it when it first started (good because
it didn't need to run for all systems, bad because that took a LONG time).
Unfortunately, I noticed that when you ran several upgrades and then
started SMC it would not work, because it would try to run some
"postrun stuff" twice and that cause breakage.
So what will happen when you lvieupgrade the same partition twice without
even once booting into it? That's easily done and should not cause
any breakage.
Casper
