Fix RPM Systemd Artifact Support #310
Conversation
frontend/rpm/template.go
Outdated
| @@ -317,6 +325,22 @@ func (w *specWrapper) PreUn() fmt.Stringer { | |||
| return b | |||
| } | |||
|
|
|||
| func systemdPostScript(unitName string, cfg dalec.SystemdUnitConfig) string { | |||
| verb := "disable" | |||
There was a problem hiding this comment.
This should take into account the default value and do nothing.
There was a problem hiding this comment.
Actually thinking about it we should probably only ever enable and do nothing when false.
I wrote up that other PR to make it a *bool with the idea that we have these preset files but I'm thinking maybe we don't need that?
basically
if !cfg.Enable {
return ""
}
// systemctl enableThere was a problem hiding this comment.
Yeah that seems reasonable. Because I believe systemctl disable is essentially a no-op unless the service is already enabled, right?
There was a problem hiding this comment.
And also potentially something we don't want to do willy-nilly.
There was a problem hiding this comment.
So this should be fixed. It does actually complicate things a bit though, because if we aren't explicitly disabling units, then we don't necessarily need a %post section or %post systemd dependency if no enabled units have been specified in the spec. I've added some logic to account for this
|
Can you add a revert of #307 since given this change I don't think we need it? |
…roperly include systemd as a %post dependency
This reverts commit 4551246.
adamperlin
force-pushed
the
adamperlin/fix-systemd-support
branch
from
fd0e806 to
5f6f703
Compare
This is done! |
What this PR does / why we need it:
Fixes #301. It is not good practice to include
.presetfiles in an RPM package so this PR changes the%postsection of dalec-generated RPM Specs to enable/disable systemd units manually.Additionally, I discovered in the course of changing this behavior that systemd was not actually marked as a
post,preunandpostundependency, meaning that prost, preun, and postun scripts using thesystemd_*macros were actually broken to begin with. This PR fixes that issue as well.Which issue(s) this PR fixes (optional, using
fixes #<issue number>(, fixes #<issue_number>, ...)format, will close the issue(s) when the PR gets merged):Fixes #
Special notes for your reviewer: