🌱 add ClusterObjectSet pre-apply dry-run validation e2e scenarios

test/e2e/features/cos_preflight.feature

@@ -0,0 +1,108 @@
+@BoxcutterRuntime
+Feature: ClusterObjectSet preflight dry-run validation
+
+  As an OLM user applying a ClusterObjectSet directly, I want the revision to be
+  rejected with an actionable PreconditionsNotMet condition when one of its
+  objects would be rejected by the API server, and to recover automatically once
+  the underlying problem is resolved, so that transient or fixable issues do not
+  require re-creating the revision.
+
+  Background:
+    Given OLM is available
+
+  Scenario: ClusterObjectSet reports PreconditionsNotMet when an object targets a non-existent namespace
+    When ClusterObjectSet is applied
+      """
+      apiVersion: olm.operatorframework.io/v1
+      kind: ClusterObjectSet
+      metadata:
+        name: ${COS_NAME}
+      spec:
+        lifecycleState: Active
+        collisionProtection: Prevent
+        group: ${NAME}
+        phases:
+        - name: resources
+          objects:
+          - object:
+              apiVersion: v1
+              kind: ConfigMap
+              metadata:
+                name: cm-x
+                namespace: ${COS_NAME}-absent
+              data:
+                k: v
+        revision: 1
+      """
+    Then ClusterObjectSet "${COS_NAME}" reports Ready as False with Reason PreconditionsNotMet and Message includes:
+    """
+      check rejected phase "resources": phase "resources": /v1, Kind=ConfigMap ${COS_NAME}-absent/cm-x: namespaces "${COS_NAME}-absent" not found.
+    """
+
+  Scenario: ClusterObjectSet reports PreconditionsNotMet when an object has an invalid spec
+    Given namespace "${TEST_NAMESPACE}" is available
+    When ClusterObjectSet is applied
+      """
+      apiVersion: olm.operatorframework.io/v1
+      kind: ClusterObjectSet
+      metadata:
+        name: ${COS_NAME}
+      spec:
+        lifecycleState: Active
+        collisionProtection: Prevent
+        group: ${NAME}
+        phases:
+        - name: resources
+          objects:
+          - object:
+              apiVersion: v1
+              kind: Pod
+              metadata:
+                name: bad-pod
+                namespace: ${TEST_NAMESPACE}
+              spec:
+                containers: []
+        revision: 1
+      """
+    Then ClusterObjectSet "${COS_NAME}" reports Ready as False with Reason PreconditionsNotMet and Message includes:
+    """
+      check rejected phase "resources": phase "resources": /v1, Kind=Pod ${TEST_NAMESPACE}/bad-pod: Pod "bad-pod" is invalid: spec.containers: Required value.
+    """
+
+  Scenario: ClusterObjectSet recovers once a missing target namespace is created
+    When ClusterObjectSet is applied
+      """
+      apiVersion: olm.operatorframework.io/v1
+      kind: ClusterObjectSet
+      metadata:
+        name: ${COS_NAME}
+      spec:
+        lifecycleState: Active
+        collisionProtection: Prevent
+        group: ${NAME}
+        phases:
+        - name: resources
+          objects:
+          - object:
+              apiVersion: v1
+              kind: ConfigMap
+              metadata:
+                name: cm-x
+                namespace: ${COS_NAME}-recover
+              data:
+                k: v
+        revision: 1
+      """
+    Then ClusterObjectSet "${COS_NAME}" reports Ready as False with Reason PreconditionsNotMet and Message includes:
+    """
+      check rejected phase "resources": phase "resources": /v1, Kind=ConfigMap ${COS_NAME}-recover/cm-x: namespaces "${COS_NAME}-recover" not found.
+    """
+    When resource is applied
+      """
+      apiVersion: v1
+      kind: Namespace
+      metadata:
+        name: ${COS_NAME}-recover
+      """
+    And ClusterObjectSet "${COS_NAME}" reconciliation is triggered
+    Then ClusterObjectSet "${COS_NAME}" reports Ready as True with Reason AllObjectsReady