Use teardown function hook

Using `teardown` function hook which will be called by scripts in
knative/test-infra to:
1. Destroy the existing environment when using an existing cluster (so
   we must wait for it to be deleted before attempting to deploy it
   again)
2. Bring down the environment after the tests complete

Also adding deployment/teardown of Build CRD which is required for the
PipelineCRD (Tasks wrap Builds)

DEVELOPMENT.md

@@ -89,10 +89,9 @@ kubectl create clusterrolebinding cluster-admin-binding \
 
 ## Deploy Knative Build
 
-build-pipeline has a dependency on [build](https://github.com/knative/build)
 
 ```
-kubectl deploy -f ./third_party/config/build/release.yaml
+kubectl apply -f ./third_party/config/build/release.yaml
 ```
 
 ## Iterating
@@ -116,11 +115,12 @@ To make changes to these CRDs, you will probably interact with:
 
 ## Running everything
 
-You can stand up a version of this controller on-cluster (to your `kubectl config current-context`) with:
+You can stand up a version of this controller on-cluster (to your `kubectl config current-context`),
+including `knative/build` (which is wrapped by [`Task`](README.md#task)):
 
 ```shell
-# This will register the CRD and deploy the controller to start acting on them.
 ko apply -f config/
+kubectl deploy -f ./third_party/config/build/release.yaml
 ```
 
 As you make changes to the code, you can redeploy your controller with:
@@ -133,7 +133,9 @@ You can clean up everything with:
 
 ```shell
 ko delete -f config/
+kubectl delete -f ./third_party/config/build/release.yaml
 ```
+
 ## Accessing logs
 
 To look at the controller logs, run:

test/e2e-tests.sh

@@ -36,13 +36,16 @@ if ! [[ -z ${KO_DOCKER_REPO} ]]; then
     export DOCKER_REPO_OVERRIDE=${KO_DOCKER_REPO}
 fi
 
-function take_down_pipeline() {
+function teardown() {
     header "Tearing down Pipeline CRD"
     ko delete --ignore-not-found=true -f config/
+    kubectl delete --ignore-not-found=true -f ./third_party/config/build/release.yaml
+    # teardown will be called when run against an existing cluster to cleanup before
+    # continuing, so we must wait for the cleanup to complete or the subsequent attempt
+    # to deploy to the same namespace will fail
+    wait_until_object_does_not_exist namespace knative-build-pipeline
+    wait_until_object_does_not_exist namespace knative-build
 }
-# TODO: add a wait for resources to be up and change take_down_pipeline to
-# teardown so that it will be called auto-magically.
-trap take_down_pipeline EXIT
 
 # Called by `fail_test` (provided by `e2e-tests.sh`) to dump info on test failure
 function dump_extra_cluster_state() {
@@ -69,16 +72,19 @@ if [[ -z ${KO_DOCKER_REPO} ]]; then
     export KO_DOCKER_REPO=${DOCKER_REPO_OVERRIDE}
 fi
 
-# Deploy the latest version of the Pipeline CRD.
-# TODO(#59) do we need to deploy the Build CRD as well?
+header "Deploying Build CRD"
+kubectl apply -f ./third_party/config/build/release.yaml
+
 header "Deploying Pipeline CRD"
 ko apply -f config/
 
-# Wait for pods to be running in the namespaces we are deploying to
+# The functions we are calling out to get pretty noisy when tracing is on
 set +o xtrace
+
+# Wait for pods to be running in the namespaces we are deploying to
 wait_until_pods_running knative-build-pipeline || fail_test "Pipeline CRD did not come up"
-set -o xtrace
 
+# Actually run the tests
 report_go_test \
 -v -tags=e2e -count=1 -timeout=20m ./test \
 ${options} || fail_test