Update for https://github.com/containers/image/pull/1299 + update tests

[mtrmac]

Does this PR introduce a user-facing change?

None

Update c/image from a WIP containers/image#1299:

• to eventually benefit from the better error messages
• for now, to exercise that PR to find issues, if any
• for now, to see if any Podman tests need adjusting.

[openshift-ci]

[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

Needs approval from an approver in each of these files:

• OWNERS [mtrmac]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[vrothberg]

Green and happy tests ✔️

[mtrmac]

c/image has now been updated, and vendoring that will break Podman tests.

So, please review/merge this PR, so that Podman tests match c/image and future vendor dances don’t stumble on the test failures.

[rhatdan]

/lgtm
/hold

[vrothberg]

Probably needs a manual rebase.

[rhatdan]

@mtrmac please rebase and repush so we can merge.

[edsantiago]

I find the new error message ("checking whether blah blah") less helpful - I think "unauthorized" was a nice useful part of the message. But anyhow, since you need to repush, could you include the "checking" part?

    is "$output" ".* checking whether a blob .* exists in localhost:${PODMAN_LOGIN_REGISTRY_PORT}/badpush: authentication required"

Reason: just plain "authentication required" is a little too generic for my tastes; it could come from other places in the code, and mask a real error. The tighter our checks, the more comfortable I feel. (And, yes, the cost is that slight error-message changes, like this one, are a hassle. I think that's a Good Thing. I don't think error messages should change often).

[mtrmac]

Good idea, updated the expected error text.

---

(Pedantically, “unauthorized” is not quite correct. This is “the user needs to log in — and after log in, the operation might well be authorized”. That’s, supposedly, different from “the user is logged in and not allowed to do this”, which now fails with requested access to the resource is denied (and previously denied: requested …).

Now, seeing how real-world registries implement the error reporting part of the spec, I have no hope of expecting these finer distinctions to be correctly distinguished.)

---

containers/image#1299 basically sets the precedent that whatever the Go error, even if it comes from a third-party library, we can and sometimes do, override the user-visible text. So improving the user-visible text in various situations is very much on the table — but it’s fairly unbounded in scope and I have other WIP work, so I’d like to leave that for someone to else to lead.

[rhatdan]

/lgtm
/hold cancel