ci(test): Separate unit tests so they don't rely on integration tests (#910)

Author: BYK

.github/workflows/test.yml

@@ -14,7 +14,18 @@ defaults:
   run:
     shell: bash
 jobs:
-  test:
+  unit-test:
+    runs-on: ubuntu-18.04
+    name: "unit tests"
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v2
+
+      - name: Unit Tests
+        working-directory: install
+        run: find ./ -type f -name "*-test.sh" -exec "./{}" \;
+
+  integration-test:
     runs-on: ubuntu-18.04
     name: "test"
     steps:
@@ -39,13 +50,6 @@ jobs:
           ./install.sh --minimize-downtime
           ./test.sh
 
-      - name: Unit Tests
-        working-directory: install
-        run: |
-          ./create-docker-volumes-test.sh
-          ./relay-credentials-test.sh
-          ./geoip-test.sh
-
       - name: Inspect failure
         if: failure()
         run: |

install/_test_setup.sh

@@ -18,7 +18,7 @@ setup() {
   # Clone the local repo into a temp dir. FWIW `git clone --local` breaks for
   # me because it depends on hard-linking, which doesn't work across devices,
   # and I happen to have my workspace and /tmp on separate devices.
-  git clone --depth=1 "file://$(pwd)" "$_SANDBOX"
+  git -c advice.detachedHead=false clone --depth=1 "file://$(pwd)" "$_SANDBOX"
 
   # Now propagate any local changes from the working copy to the sandbox. This
   # provides a pretty nice dev experience: edit the files in the working copy,