#909 Using a non-empty variant for ostree_installer errors without buildinstall
Closed: Fixed 5 years ago Opened 6 years ago by puiterwijk.

I have a pungi config that explicitly marks "buildinstall" as a skipped step, but the ostree_installer configuration still fails to verify because the ostree_installer and buildinstall phases would conflict.
Maybe it's an idea to change the check in pungi/phases/ostree_installer.py to if conf and not variant.is_empty and 'buildinstall' not in self.compose.skip_phases, or would that still conflict with another phase?


Me too.
I have discovered the same class of issue with my compose. The expectation is ostree_installer replaces the normal buildinstall, and it's my opinion that it's acceptable for ostree_installer to clobber or overlay any existing things within the scope of the variant. At least that is the situation in my use-case. For now I'm trying to work around this with an side variant for ostree_installer.

We could turn the error into a warning. That should solve the problem that @parasense is having, but such warning is really easy to miss and it it would basically bring #695 back.

Commit 7c237c2 relates to this ticket

It looks like 543154d doesn't quite work. The _skipped property of the buildinstall phase doesn't seem to be set quite right, see: https://kojipkgs.fedoraproject.org/compose/iot/Fedora-IoT-28-20180524.2/logs/global/.

The ostree_installer_overwrite option seems to work (at least allows the compose to start), so I'm using that as a workaround for now.

Metadata Update from @puiterwijk:
- Issue status updated to: Open (was: Closed)

5 years ago

Ah, I see why: ._skipped is only set by start(): https://pagure.io/pungi/blob/master/f/pungi/phases/base.py#_57

We should probably change the check in the ostree_installer phase to use self.bi.skip().

We should probably change the check in the ostree_installer phase to use self.bi.skip().

That will fix it, but in case buildinstall phase is not running because the product is not bootable there will be duplicated log message. Not a big deal I guess.

Login to comment on this ticket.

Metadata
Related Pull Requests
  • #960 Merged 5 years ago
  • #937 Merged 5 years ago