Using crd-diff to determine breaking changes

[rashmi43]

Non deterministic output as this is not helpful and can we consider returning a result flag that evaluates to true or false. Using crd-diff to compare EDB operator release versions

2025/04/10 20:57:58 comparing the CRDs identified incompatible changes

validations failed:
	comparing same versions:
		comparing version "v1" with version "v1":
			comparing property "^.spec.replica.primary":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.monitoring.tls":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.externalClusters":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.postInitSQLRefs":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.replica.minApplyDelay":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.managed.services":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.locale":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.imageCatalogRef":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.plugins":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.replica.self":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.postInitTemplateSQLRefs":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.managed.services.additional":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.import.pgDumpExtraOptions":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.import.pgRestoreExtraOptions":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.localeProvider":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.replicationSlots.synchronizeReplicas":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.icuRules":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.replica.promotionToken":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.builtinLocale":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.postgresql.synchronous":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.bootstrap.initdb.icuLocale":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.managed.services.additional[*].serviceTemplate.metadata.name":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.externalClusters[*].plugin":
				unknown change(s), refusing to determine that change is safe

			comparing property "^.spec.serviceAccountTemplate.metadata.name":
				unknown change(s), refusing to determine that change is safe

[everettraven]

In the latest release there were some pretty significant changes, but there still isn't explicit determinism in the output. I think this is reasonable to add for things like the plaintext output format, but I think you'd be hard pressed for determinism in output formats like JSON/YAML with the more dynamic nature of the output and not having control over the order in which it is serialized.

As for "returning a flag that evaluates to true or false", I'm not sure what you are referring to here. Could you elaborate on what you are looking for?