Re: [devel] [PATCH for rpm] Add triggers circumvension for packagekit offline update

["Dmitry V. Levin" <ldv@altlinux.org>]

[-- Attachment #1: Type: text/plain, Size: 1731 bytes --]

On Thu, Oct 31, 2019 at 04:19:16PM +0300, Aleksei Nikiforov wrote:
> 31.10.2019 15:59, Dmitry V. Levin пишет:
> > On Thu, Oct 31, 2019 at 02:58:08PM +0300, Aleksei Nikiforov wrote:
> >> Joint change with Oleg Solovyov (mcpain@)
> > 
> > Thanks.
> > 
> > It's not clear from the commit message why one might need to disable
> > rpmdb --rebuilddb after rpm update.
> > 
> > Could you elaborate, please?
> 
> packagekit offline update is implemented in following way:
> 1) system is preparing for offline update: all new installed packages 
> are downloaded, etc.
> 2) system is rebooting into special systemd offline update mode upon 
> user request.
> 3) in this offline update mode packagekit offline update service is 
> performing system update using packagekit service activated via dbus.
> 4) Last step of systemd offline update is a reboot request. It must be 
> made by offline update service. And packagekit offline update service 
> does request it.
> 
> Thus, rpmdb --rebuilddb from rpm scripts would be in race against reboot 
> request. Testing showed that it doesn't work properly in such 
> conditions. In best case, rpmdb --rebuilddb doesn't do anything at all 
> before reboot happens. Instead of hoping to get it work properly in such 
> conditions, disabling it in rpm scripts and calling it right after 
> update but before reboot request from packagekit offline update works fine:

OK, I cannot say that I like this approach, but at least it makes sense.
The alternative is invoking pk-offline-update in a separate pid namespace
which is probably not as simple as disabling rpm %post.

Could you add the rationale from your post into the commit message,
please?


-- 
ldv

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 801 bytes --]