From 2820cd34760007c054c04e8f9bd6f27f10be288e Mon Sep 17 00:00:00 2001
From: Edwin van den Houdt <eho@respellion.nl>
Date: Tue, 21 Jul 2026 20:07:35 +0200
Subject: [PATCH] =?UTF-8?q?fix(ci):=20e2e=20=E2=80=94=20let=20Playwright?=
 =?UTF-8?q?=20manage=20both=20servers=20(fixes=20the=202-hour=20hang)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The e2e job backgrounded `dotnet run &` and `ng serve &` in separate Actions
steps, then `npx wait-on` (no timeout) in a later step. A process started with
`&` in one step is killed when that step's shell exits, so wait-on waited forever
on servers that were already gone — the job hung until the runner's hard limit
(~2h; Gitea's act_runner doesn't reliably enforce timeout-minutes).

Move both servers into Playwright's `webServer` (an array: backend + `npm start`),
so Playwright starts them, waits for readiness, runs the suite, and tears them
down in the one `npm run e2e` process — CI and local alike. The CI e2e job is now
just npm ci / playwright install / npm run e2e. `reuseExistingServer` is on locally
(reuses a running app, incl. the docker stack) and off in CI (fresh start).

Verified locally via `CI=1 npm run e2e` against free ports + a clean db: both
smoke tests pass in ~15s (no hang). The earlier local failure was leftover SQLite
state (WP-22) from a dirty run resuming a Concept — CI checks out fresh, so its db
is always empty.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---
 .github/workflows/ci.yml |  8 +++++---
 playwright.config.ts     | 33 ++++++++++++++++++++++-----------
 scripts/ci-local.sh      |  7 ++++---
 3 files changed, 31 insertions(+), 17 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 424a3d2..8f9052e 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -68,6 +68,11 @@ jobs:
     # runner checkout per run, so there's no bigregister.db (WP-22, gitignored)
     # left over from a prior run to leak state in; the backend creates + migrates
     # an empty one on this boot, same as a fresh clone always has.
+    # Playwright's `webServer` (playwright.config.ts) starts BOTH the backend and
+    # `ng serve`, waits for them, runs the suite, and tears them down — all in the
+    # one `npm run e2e` process. Do NOT background them as separate steps: a `&`
+    # process from one Actions step is dead by the next step, so `wait-on` hung
+    # forever (the 2-hour e2e hang).
     runs-on: ubuntu-latest
     timeout-minutes: 15
     steps:
@@ -81,9 +86,6 @@ jobs:
           dotnet-version: 10.0.x
       - run: npm ci
       - run: npx playwright install --with-deps chromium
-      - run: dotnet run --project backend/src/BigRegister.Api --urls http://localhost:5000 &
-      - run: npx ng serve --proxy-config proxy.conf.json &
-      - run: npx wait-on http://localhost:5000/swagger http://localhost:4200
       - run: npm run e2e
 
   codeql:
diff --git a/playwright.config.ts b/playwright.config.ts
index ef1f86c..0106bc7 100644
--- a/playwright.config.ts
+++ b/playwright.config.ts
@@ -15,15 +15,26 @@ export default defineConfig({
     trace: 'on-first-retry',
   },
   projects: [{ name: 'chromium', use: { browserName: 'chromium' } }],
-  // CI starts `ng serve` + the backend as separate job steps (both need to be up
-  // before the suite runs); locally, boot `npm start` automatically so `npm run e2e`
-  // works standalone — the backend still needs `dotnet run` running separately.
-  webServer: process.env['CI']
-    ? undefined
-    : {
-        command: 'npm start',
-        url: baseURL,
-        reuseExistingServer: true,
-        timeout: 120_000,
-      },
+  // Playwright owns both servers' lifecycle — start, wait-for-ready, tear down — in the
+  // one `npm run e2e` process, so `npm run e2e` is self-contained locally AND in CI.
+  // Do NOT background them as separate CI steps: a process started with `&` in one
+  // Actions `run:` step is killed when that step's shell exits, so a later `wait-on`
+  // step hangs forever on servers that are already gone (the 2-hour e2e hang).
+  // `reuseExistingServer` locally lets you run against an already-running app (incl. the
+  // docker stack on 4200/5000); CI always starts fresh. Backend gets a longer timeout —
+  // `dotnet run` cold-restores+builds before it listens.
+  webServer: [
+    {
+      command: 'dotnet run --project backend/src/BigRegister.Api --urls http://localhost:5000',
+      url: 'http://localhost:5000/swagger',
+      reuseExistingServer: !process.env['CI'],
+      timeout: 180_000,
+    },
+    {
+      command: 'npm start',
+      url: baseURL,
+      reuseExistingServer: !process.env['CI'],
+      timeout: 120_000,
+    },
+  ],
 });
diff --git a/scripts/ci-local.sh b/scripts/ci-local.sh
index e9527bf..b93d09f 100755
--- a/scripts/ci-local.sh
+++ b/scripts/ci-local.sh
@@ -25,6 +25,7 @@ if [[ "${1:-}" == "--full" ]]; then
 fi
 
 printf '\n\033[1;32m✔ local CI passed\033[0m\n'
-printf 'Note: the e2e job is not run here — it needs the backend + ng serve up. Run:\n'
-printf '  (backend) dotnet run --project backend/src/BigRegister.Api --urls http://localhost:5000 &\n'
-printf '  (frontend) npx ng serve &  then  npm run e2e\n'
+printf 'Note: the e2e job is not chained here (it is slow). Run it standalone with:\n'
+printf '  npm run e2e\n'
+printf 'Playwright starts the backend + ng serve itself (playwright.config.ts webServer),\n'
+printf 'reusing an already-running app on 4200/5000 if present.\n'