Am Wed, 21 Jul 2021 06:00:12 -0700 schrieb Andrea Bolognani <abologna@xxxxxxxxxx>: > On Wed, Jul 21, 2021 at 01:23:56PM +0200, Olaf Hering wrote: > > There is no point in restarting libvirtd in the middle of the transaction. > > The spec file gives no ordering hints to rpm. > Note that the daemon restart always happens *after* the transaction, > in the %posttrans part. I was wrong, %systemd_post may not do a 'try-restart', but that certainly depends on what this macro actually does. At least the macros I have will just do 'daemon-reload', they do nothing with the listed units. It is not clear if 'try-restart' reloads the EnvironmentFile=. If it is indeed reloaded, a modified file will not exist at this point because it was just renamed to .rpmsave. In other words: the current '%post daemon' needs no modification. > It looks like the only documentation for the various options that can > be used to tweak the service's behavior was the one found in the > sysconfig file. Perhaps a man page should be introduced? Yeah, someone familiar with libvirt-guests.sh should provide one. This patch just moves the existing comments elsewhere. To mean it does not look like a user needs to tweak that script. > But then libvirt-guests would end up being the only service that > cannot be configured via a systemd override... I think we should > strive for consistency here. It certainly can be configured. The environment can be constructed either with a .service.d/ file or with the (optional) sysconfig file. Olaf
Attachment:
pgpi69wmnSQE1.pgp
Description: Digitale Signatur von OpenPGP
