Revert "compose: Inject our static tmpfiles.d dropins earlier" #4881
Conversation
This reverts commit e1e78cf. It breaks idempotency with osbuild. Closes: coreos#4879
|
Didn't test this locally; presumably this will break the stateoverlay testing, which we would also probably need to unwind. |
|
Yeah, you probably need to revert fde9e9e as well |
|
I tested practically the same change and it seemed to fix things, but just in case I'm running a build on this branch locally one more time |
I kind of hope so yes...yeah, might as well go ahead and do that too now |
|
I think the actual fix for this is to make the |
This reverts commit a5046b7.
cgwalters
force-pushed
the
postprocess-idempotent
branch
from
2d720a0 to
8f176e8
Compare
Yeah...I started looking/thinking about this but the complexity is really high, and we probably need to add the osbuild testing to have some confidence right? |
Yup, sure. I still think we should revert for now. |
|
This passed the AutoSD build locally on my machine |
|
But actually, as per what we discussed in ostreedev/ostree-rs-ext#607 (comment), we don't need to support turning it on in base composes anyway. I'll open a follow up to this that removes it. |
|
Can we also get a new fedora/cs9 release with these changes? Either as backported patches or a new release. |
This comment was marked as outdated.
This comment was marked as outdated.
This reverts commit e1e78cf.
It breaks idempotency with osbuild.
Closes: #4879