[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: mtrmac

The full list of commands accepted by this bot can be found here.

The pull request process is described here

@Luap99 @edsantiago Here I see tests failing on

[FAIL] Podman run [It] podman run and decrypt from local registry
/var/tmp/go/src/github.com/containers/podman/test/e2e/run_test.go:2187
[FAIL] Podman pull podman pull and decrypt [It] From oci
/var/tmp/go/src/github.com/containers/podman/test/e2e/pull_test.go:618

where an image is pushed encrypted, removed locally, and then we test that a pull without an encryption key fails.

It is the kind of thing that might well be broken by the work I am testing, but I don’t see this failing when testing locally, and I can’t currently see a code path that could cause an unexpected reuse.

Is it possible that this is newly being triggered by a race in tests? I’m afraid I’m only vaguely aware of the test changes in Podman.

Is it possible that this is newly being triggered by a race in tests?

Trying now with more unique image names, which might explain this.

Like you, I'm unable to reproduce locally. Even with CI_USE_REGISTRY_CACHE=1 (which was a long shot anyway).

For reference, this is the (correct) output from a recent run on another PR:

           Trying to pull localhost:5006/my-alpine:latest...
           Getting image source signatures
           Copying blob sha256:da4dd051a6bdde5e43f55eafc0cd715657f54b80f88af6742a53971eb71e4a39
           Error: writing blob: adding layer with blob "sha256:da4dd051a6bdde5e43f55eafc0cd715657f54b80f88af6742a53971eb71e4a39": unpacking failed (error: exit status 1; output: archive/tar: invalid tar header)
           exhausting input failed (error: archive/tar: invalid tar header)

(that is also what I see on my f40 laptop).

And this is what we see in CI for this PR:

           Trying to pull localhost:5006/my-alpine-podman-run-and-decrypt:latest...
           Getting image source signatures
           Copying blob sha256:937f71f790005574ab9b667290dbfe40df78b0b200fa42631e755f212ddd73be
           Copying config sha256:4819930274ea2bd6ab611972f9928e07b84edcbfe1d2462fb2941957308dfae2
           Writing manifest to image destination

           [FAILED] Failure recorded during attempt 1:
           Command exited with status 0 (expected 125)

I'm confused too.

I’m sorry, this is actually directly related to the c/image changes I’m testing. It does not reproduce in isolation, the failure happens because the BlobInfoCache has a record of a previous, unrelated, push of the same image.

Thank you for the followup!

I have split the Podman-side test fixes into #23379; this now tests the combination of #23379 with the c/image PR.

containers/image#2321 was merged.