Conversion webhooks in main

.github/workflows/dockerimage-next.yml

@@ -46,7 +46,7 @@ jobs:
       DWO_BUNDLE_IMG: quay.io/devfile/devworkspace-operator-bundle:next
       DWO_INDEX_IMG: quay.io/devfile/devworkspace-operator-index:next
       OPM_VERSION: v1.17.1
-      OPERATOR_SDK_VERSION: v1.7.2
+      OPERATOR_SDK_VERSION: v1.8.0
     steps:
 
       - name: Cache Operator SDK ${{ env.OPERATOR_SDK_VERSION }}

Makefile

@@ -45,7 +45,7 @@ ifneq (,$(shell which kubectl 2>/dev/null)$(shell which oc 2>/dev/null))
 include build/make/deploy.mk
 endif
 
-OPERATOR_SDK_VERSION = v1.7.2
+OPERATOR_SDK_VERSION = v1.8.0
 OPM_VERSION = v1.17.1
 
 # Produce CRDs that work back to Kubernetes 1.11 (no version conversion)

build/make/olm.mk

@@ -20,6 +20,13 @@ generate_olm_bundle_yaml: _check_operator_sdk_version _generate_olm_deployment_f
 		--channels fast \
 		--metadata && \
 	mv bundle.Dockerfile build/
+# Operator SDK v1.8.0 does not output webhooks in a stable order, so we have to sort the yaml files to avoid
+# spurious changes. See issue https://github.com/operator-framework/operator-sdk/issues/5022
+	yq -iY '.spec.webhookdefinitions |= sort' deploy/bundle/manifests/devworkspace-operator.clusterserviceversion.yaml
+# OLM creates a configmap that contains the files in bundle when an operator is installed. Since the maximum size
+# of a resource in etcd is 1MiB, we need to do a bit of squishing on our yaml files to get the total bundle under the
+# 1MiB limit. This command puts all YAML strings on a single line, avoiding ~200KiB of newlines and indentation.
+	find deploy/bundle/manifests -name '*.yaml' -exec yq --indentless -w 1000000000 -iY . {} \;
 
 ### build_bundle_image: build and push DevWorkspace Operator bundle image
 build_bundle_image: _print_vars _check_operator_sdk_version