[release-5.31] Don't abort listing tags when we encounter a digest

[TomSweeneyRedHat]

Cherry Pick #2440 then bump to v5.31.1 and finally to v5.31.2-dev

[TomSweeneyRedHat]

@mtrmac Do I need to adjust the test, or is this picking up a true error? I see the same in the 5.29 branch, but the 5.30 branch is having a compile error, and I'm confused as to why 5.29 isn't, too.

        utils.go:70: Running skopeo sync --scoped --src yaml --dest dir /tmp/TestSyncTestYamlUntagged3473006696/001/registries.yaml /tmp/TestSyncTestYamlUntagged3473006696/001/dir1
        sync_test.go:316: 
            	Error Trace:	/var/tmp/go/src/github.com/containers/skopeo/integration/sync_test.go:316
            	Error:      	"[v1.6.6 sha256-017727efcfeb7d053af68e51436ce8e65edbc6ca573720afb4f79c8594036955.sig sha256-8e352a029d304ca7431c6507b56800636c321cb52289686a581ab70aaa8a2e2a.sig sha256-a0ead06651cf580044aeb0a0feba63591858fb2e43ade8c9dea45a6a89ae7e5e.sig sha256-b82e294de6be763f73ae71266c8f5466e7e03c69f3a1de96efd570284d35bb18.sig v1.10.0 v1.10.1]" should have 18 item(s), but has 7
            	Test:       	TestSync/TestYamlUntagged
        sync_test.go:127: 
            	Error Trace:	/var/tmp/go/src/github.com/containers/skopeo/integration/sync_test.go:127
            	            				/var/tmp/go/src/github.com/containers/skopeo/integration/sync_test.go:317
            	Error:      	Not equal: 
            	            	expected: 18
            	            	actual  : 7
            	Test:       	TestSync/TestYamlUntagged

[mtrmac]

@TomSweeneyRedHat Thanks for the backport!

The root cause of the count discrepancy is a failure a few lines above:

… msg="Copying image ref 7/18" …
… msg="Error copying ref \"docker://k8s.gcr.io/coredns/coredns:v1.6.6\": initializing source docker://k8s.gcr.io/coredns/coredns:v1.6.6: invalid character '<' looking for beginning of value"

Sadly that failure does not immediately abort the test. IIRC we’ve been seeing k8s.gcr.io failures all over the place in the past few days, I’m not sure whether that has been resolved.

---

LGTM other than the test failure.

[mtrmac]

(See a podman-monitor thread about the test failure; also #2454 .)

[TomSweeneyRedHat]

@Luap99 you were concerned that this might be a breaking change? Can you confirm either in here or Slack and let me know what's needed. We have a high level customer request to fix this issue quickly.

@cevich or @edsantiago any thoughts on the testing hiccups?

[Luap99]

@TomSweeneyRedHat I was talking about containers/storage#1828 not anything in c/image. If you backported the c/storage PR without the follow up revert on the broken commit we are breaking backwards compatibility on all podman containers. But anyway it is in now way related to this here so lets have this discussion elsewhere.

The k8s.gcr.io failures must be fixed by replacing the URL to new registry AFAICT, i.e. containers/common#2057 and containers/skopeo#2356

[TomSweeneyRedHat]

This is a backport to begin the fix for containers/skopeo#2346 in the versions of container tools that use c/image v5.31.