Skip to content
E2E Tests

Add grading assignment defaults and reminder automation #3259

Sign in to view logs

Add grading assignment defaults and reminder automation

Add grading assignment defaults and reminder automation #3259

Workflow file for this run

.github/workflows/deploy.yml at 3e1ae94
name: E2E Tests
on:
push:
branches:
- main
- staging
pull_request_target:
workflow_dispatch:
# Secure baseline: no permissions. Jobs that need them override explicitly.
permissions: {}
# Cancel an in-flight run when a newer one comes in for the same PR (or push
# ref). Under pull_request_target, github.ref resolves to the base branch, so
# keying on it alone would make every PR against staging collide; prefer
# pull_request.number when present so each PR has its own group.
concurrency:
group: e2e-local-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
jobs:
# Local-Supabase E2E. Spins up a clean dev environment on the runner and runs
# the Playwright suite against it. The Coolify branch-preview job that used to
# run alongside this one is gone — it'll come back once the deployment target
# moves into the same k8s cluster as this runner.
e2e-local:
runs-on: pawtograder-e2e
timeout-minutes: 75
permissions:
contents: read
env:
# Per-run project name so that two e2e-local jobs running on different
# refs (or different runners in the same pool) don't clobber each
# other's Supabase containers / Docker volumes. The Supabase CLI takes
# this from `supabase/config.toml` `project_id`, so we sed-edit the
# checkout copy below before any `supabase` invocation.
SUPABASE_PROJECT: pawtograder-platform-${{ github.run_id }}-${{ github.run_attempt }}
# Supabase local-dev keys are deterministic (signed with the public
# "supabase-demo" JWT secret). Hardcoding lets us write .env.local — and
# therefore start `next build` — before `supabase start` finishes.
LOCAL_SUPABASE_URL: http://127.0.0.1:54321
LOCAL_SUPABASE_ANON_KEY: eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJzdXBhYmFzZS1kZW1vIiwicm9sZSI6ImFub24iLCJleHAiOjE5ODM4MTI5OTZ9.CRXP1A7WOeoJeXxjNni43kdQwgnWNReilDMblYTn_I0
LOCAL_SUPABASE_SERVICE_ROLE_KEY: eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJzdXBhYmFzZS1kZW1vIiwicm9sZSI6InNlcnZpY2Vfcm9sZSIsImV4cCI6MTk4MzgxMjk5Nn0.EGIM96RAZx35lJzdJsyH-qQwv8Hdp7fsn3W0YpN81IU
# Per-run scratch dir for PID / log files. /tmp persists across jobs on
# self-hosted runners, so namespacing on run_id+attempt prevents the
# cleanup `kill $(cat ...pid)` from targeting a stale PID that the OS
# has since recycled to an unrelated process.
RUN_TMP: /tmp/e2e-${{ github.run_id }}-${{ github.run_attempt }}
steps:
- uses: actions/checkout@v4
with:
ref: ${{ github.event.pull_request.head.sha || github.sha }}
- name: Cache npm dependencies
uses: actions/cache@v4
with:
path: ~/.npm
key: npm-${{ runner.os }}-node-22-${{ hashFiles('package-lock.json') }}
restore-keys: |
npm-${{ runner.os }}-node-22-
- name: Cache Next.js build
uses: actions/cache@v4
with:
path: .next/cache
key: nextjs-${{ runner.os }}-${{ hashFiles('package-lock.json') }}-${{ hashFiles('**/*.[jt]s', '**/*.[jt]sx', '!**/node_modules/**') }}
restore-keys: |
nextjs-${{ runner.os }}-${{ hashFiles('package-lock.json') }}-
nextjs-${{ runner.os }}-
- uses: actions/setup-node@v4
with:
node-version: 22
- uses: supabase/setup-cli@v1
with:
# Pin to a known-good version. `latest` makes runs non-reproducible —
# bump deliberately when we want a new CLI.
version: 2.92.1
- name: Install dependencies
run: npm ci
- name: Write .env.local
env:
GITHUB_APP_ID: ${{ secrets.PAWTOGRADER_GITHUB_APP_ID }}
GITHUB_PRIVATE_KEY_STRING: ${{ secrets.PAWTOGRADER_GITHUB_PRIVATE_KEY_STRING }}
GITHUB_OAUTH_CLIENT_ID: ${{ secrets.PAWTOGRADER_GITHUB_OAUTH_CLIENT_ID }}
GITHUB_OAUTH_CLIENT_SECRET: ${{ secrets.PAWTOGRADER_GITHUB_OAUTH_CLIENT_SECRET }}
UPSTASH_REDIS_REST_URL: ${{ secrets.UPSTASH_REDIS_REST_URL }}
UPSTASH_REDIS_REST_TOKEN: ${{ secrets.UPSTASH_REDIS_REST_TOKEN }}
SENTRY_DSN: ${{ secrets.SENTRY_DSN }}
run: |
set -euo pipefail
cat > .env.local <<EOF
NEXT_PUBLIC_SUPABASE_URL=${LOCAL_SUPABASE_URL}
NEXT_PUBLIC_SUPABASE_ANON_KEY=${LOCAL_SUPABASE_ANON_KEY}
SUPABASE_URL=${LOCAL_SUPABASE_URL}
SUPABASE_ANON_KEY=${LOCAL_SUPABASE_ANON_KEY}
SUPABASE_SERVICE_ROLE_KEY=${LOCAL_SUPABASE_SERVICE_ROLE_KEY}
ENABLE_SIGNUPS=true
NEXT_PUBLIC_PAWTOGRADER_WEB_URL=http://localhost:3001
E2E_ENABLE=true
END_TO_END_SECRET=not-a-secret
EDGE_FUNCTION_SECRET=some-secret-value
GITHUB_APP_ID=${GITHUB_APP_ID}
GITHUB_OAUTH_CLIENT_ID=${GITHUB_OAUTH_CLIENT_ID}
GITHUB_OAUTH_CLIENT_SECRET=${GITHUB_OAUTH_CLIENT_SECRET}
UPSTASH_REDIS_REST_URL=${UPSTASH_REDIS_REST_URL}
UPSTASH_REDIS_REST_TOKEN=${UPSTASH_REDIS_REST_TOKEN}
SENTRY_DSN=${SENTRY_DSN}
EOF
# Single-line escaped PEM — supabase functions serve --env-file does
# not support multi-line quoted values; .env.local.staging uses the
# same shape. The Octokit / GitHub-App SDK unescapes \n at runtime.
KEY_ESCAPED=$(printf '%s' "${GITHUB_PRIVATE_KEY_STRING}" | awk 'BEGIN{ORS=""} {printf "%s\\n", $0}')
printf 'GITHUB_PRIVATE_KEY_STRING="%s"\n' "$KEY_ESCAPED" >> .env.local
# Make supabase/config.toml's project_id unique per run so the supabase
# CLI tags its containers and volumes with our SUPABASE_PROJECT, not the
# repo-default `pawtograder-platform`. Without this, two concurrent runs
# would mutually clobber each other's stack.
- name: Pin supabase project_id per run
run: |
set -euo pipefail
sed -i.bak -E "s|^project_id = .*$|project_id = \"${SUPABASE_PROJECT}\"|" supabase/config.toml
grep "^project_id" supabase/config.toml
# Kick Supabase off in the background so its container pulls + migration
# run overlap with `next build` and `playwright install`.
- name: Start local Supabase (background, fresh)
run: |
set -euo pipefail
mkdir -p "$RUN_TMP"
rm -f "$RUN_TMP/supabase-start.exitcode"
# `supabase start` occasionally races its own internal stop/restart and
# dies with "failed to bind host port 0.0.0.0:54322/tcp: address already
# in use" — a container from the in-progress attempt hasn't released the
# fixed port before the next bind. Reset this run's containers/volumes and
# retry. The reset is scoped to SUPABASE_PROJECT (unique per run_id +
# attempt), so concurrent jobs sharing the runner pool are never touched.
# Backgrounded so the container pulls + migrations overlap with the build.
nohup bash -c '
set -uo pipefail
reset_supabase() {
npx supabase stop --no-backup >/dev/null 2>&1 || true
docker ps -aq --filter "label=com.supabase.cli.project=${SUPABASE_PROJECT}" \
| xargs -r docker rm -f >/dev/null 2>&1 || true
docker volume ls --filter "label=com.supabase.cli.project=${SUPABASE_PROJECT}" -q \
| xargs -r docker volume rm >/dev/null 2>&1 || true
}
reset_supabase
for attempt in $(seq 1 5); do
echo "=== supabase start attempt ${attempt}/5 ==="
if npx supabase start; then
echo 0 > "${RUN_TMP}/supabase-start.exitcode"
exit 0
fi
echo "supabase start failed on attempt ${attempt}; resetting and retrying"
reset_supabase
sleep $((attempt * 3))
done
echo 1 > "${RUN_TMP}/supabase-start.exitcode"
exit 1
' > "$RUN_TMP/supabase-start.log" 2>&1 &
echo $! > "$RUN_TMP/supabase-start.pid"
# Background the browser install (~30-90s, mostly downloads + apt-get).
- name: Install Playwright Browsers (background)
run: |
set -euo pipefail
rm -f $RUN_TMP/playwright-install.exitcode
nohup bash -c 'npx playwright install --with-deps \
> $RUN_TMP/playwright-install.log 2>&1; \
echo $? > $RUN_TMP/playwright-install.exitcode' &
- name: Build Next.js (prod, port 3001)
env:
NEXT_PUBLIC_PAWTOGRADER_WEB_URL: http://localhost:3001
run: npm run build
- name: Wait for Playwright install
run: |
set -euo pipefail
for i in $(seq 1 600); do
[ -f $RUN_TMP/playwright-install.exitcode ] && break
sleep 1
done
rc="$(cat $RUN_TMP/playwright-install.exitcode 2>/dev/null || echo 'timeout')"
if [ "$rc" != "0" ]; then
echo "playwright install failed (rc=$rc)"
echo '--- playwright-install.log ---'
cat $RUN_TMP/playwright-install.log || true
exit 1
fi
- name: Wait for Supabase + finalize schema
run: |
set -euo pipefail
# Poll the API until it answers. The background launcher retries
# `supabase start` (with a reset between attempts) and writes its final
# status to supabase-start.exitcode; if that lands non-zero, all retries
# were exhausted, so fail fast with the log instead of polling for the
# full timeout.
for i in $(seq 1 600); do
if curl -sf -o /dev/null \
-H "apikey: ${LOCAL_SUPABASE_ANON_KEY}" \
"${LOCAL_SUPABASE_URL}/rest/v1/"; then
break
fi
if [ -f "$RUN_TMP/supabase-start.exitcode" ] \
&& [ "$(cat "$RUN_TMP/supabase-start.exitcode")" != "0" ]; then
echo 'supabase start exhausted its retries and never came up'
echo '--- supabase-start.log ---'
cat "$RUN_TMP/supabase-start.log" || true
exit 1
fi
sleep 1
done
# Final readiness check — fail loudly if we fell out of the loop.
curl -sf -o /dev/null \
-H "apikey: ${LOCAL_SUPABASE_ANON_KEY}" \
"${LOCAL_SUPABASE_URL}/rest/v1/"
# Ephemeral CI Postgres — trade durability for speed. The volume is
# destroyed at job end, so fsync/full_page_writes give us nothing.
# synchronous_commit is SIGHUP-reloadable; fsync, full_page_writes,
# and shared_buffers require a restart (done below).
# Notes on the psql invocation:
# * Each ALTER SYSTEM must run outside a transaction block, so use
# one -c flag per statement — `psql -c "a;b;c"` wraps multi-
# statement input in an implicit transaction and would fail.
# * Connect as supabase_admin (the actual superuser) — the
# `postgres` role on supabase-local is not superuser and gets
# `permission denied to set parameter` on ALTER SYSTEM.
# * Use -h 127.0.0.1 so pg_hba.conf's `host ... 127.0.0.1 trust`
# rule applies; the unix-socket rule requires scram-sha-256.
docker exec -i "supabase_db_${SUPABASE_PROJECT}" \
psql -v ON_ERROR_STOP=1 -h 127.0.0.1 -U supabase_admin -d postgres \
-c "ALTER SYSTEM SET synchronous_commit = off" \
-c "ALTER SYSTEM SET fsync = off" \
-c "ALTER SYSTEM SET full_page_writes = off" \
-c "ALTER SYSTEM SET shared_buffers = '512MB'" \
-c "ALTER SYSTEM SET work_mem = '32MB'" \
-c "ALTER SYSTEM SET maintenance_work_mem = '256MB'" \
-c "ALTER SYSTEM SET autovacuum = off"
docker restart "supabase_db_${SUPABASE_PROJECT}"
# Re-wait for REST after the db bounce — PostgREST reconnects lazily
# but Kong returns 5xx until the upstream is healthy again.
for i in $(seq 1 120); do
if curl -sf -o /dev/null \
-H "apikey: ${LOCAL_SUPABASE_ANON_KEY}" \
"${LOCAL_SUPABASE_URL}/rest/v1/"; then
break
fi
sleep 1
done
curl -sf -o /dev/null \
-H "apikey: ${LOCAL_SUPABASE_ANON_KEY}" \
"${LOCAL_SUPABASE_URL}/rest/v1/"
# Audit table is partitioned; seed today's partition (and the next 7 days).
docker exec -i "supabase_db_${SUPABASE_PROJECT}" psql -U postgres -d postgres -c \
"SELECT public.audit_maintain_partitions();"
- name: Serve Edge Functions
run: |
set -euo pipefail
nohup npx supabase functions serve --env-file .env.local \
> $RUN_TMP/edge-functions.log 2>&1 &
echo $! > $RUN_TMP/edge-functions.pid
# Wait until the gateway answers. `curl -s -w %{http_code}` prints
# `000` on connect-refused, so check the exit status (or filter
# `000`) — otherwise the loop breaks on the first tick.
ready=0
for i in $(seq 1 60); do
code=$(curl -s -o /dev/null -w '%{http_code}' http://127.0.0.1:54321/functions/v1/ || true)
if [ -n "$code" ] && [ "$code" != "000" ]; then ready=1; break; fi
sleep 1
done
if [ "$ready" != "1" ]; then
echo "Edge Functions never became reachable"
tail -n 200 $RUN_TMP/edge-functions.log || true
exit 1
fi
- name: Start Next.js prod server
run: |
set -euo pipefail
# CSP is served in report-only mode for E2E so Playwright's
# page.evaluate / waitForFunction (string-eval via CDP) is not
# blocked by `script-src` strict-dynamic. The header is still
# emitted, so tests can still assert its shape; enforcement
# ships in prod via the default `Content-Security-Policy`.
nohup env PORT=3001 CSP_REPORT_ONLY=1 npm run start > $RUN_TMP/next-server.log 2>&1 &
echo $! > $RUN_TMP/next-server.pid
ready=0
for i in $(seq 1 60); do
if curl -sf -o /dev/null http://localhost:3001/; then ready=1; break; fi
sleep 1
done
if [ "$ready" != "1" ]; then
echo "Next.js prod server never became reachable"
tail -n 200 $RUN_TMP/next-server.log || true
exit 1
fi
- name: Run Playwright tests
env:
BASE_URL: http://localhost:3001
SUPABASE_URL: ${{ env.LOCAL_SUPABASE_URL }}
SUPABASE_ANON_KEY: ${{ env.LOCAL_SUPABASE_ANON_KEY }}
SUPABASE_SERVICE_ROLE_KEY: ${{ env.LOCAL_SUPABASE_SERVICE_ROLE_KEY }}
# EDGE_FUNCTION_SECRET / END_TO_END_SECRET are sourced from .env.local
# (written above) via the test-runner's dotenv load — single source
# of truth so the two values can't drift.
TEST_PASSWORD: ${{ secrets.TEST_PASSWORD }}
ARGOS_TOKEN: ${{ secrets.ARGOS_TOKEN }}
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: npx playwright test
- name: Tail server logs on failure
if: failure()
run: |
echo '--- edge-functions.log ---'; tail -n 500 $RUN_TMP/edge-functions.log || true
echo '--- next-server.log ---'; tail -n 500 $RUN_TMP/next-server.log || true
- name: Stop background services
if: always()
run: |
kill "$(cat $RUN_TMP/edge-functions.pid 2>/dev/null)" 2>/dev/null || true
kill "$(cat $RUN_TMP/next-server.pid 2>/dev/null)" 2>/dev/null || true
# Capture full Supabase container logs while the stack is still up.
mkdir -p server-logs
for c in $(docker ps --filter "label=com.supabase.cli.project=${SUPABASE_PROJECT}" --format '{{.Names}}'); do
docker logs "$c" > "server-logs/${c}.log" 2>&1 || true
done
cp -f $RUN_TMP/edge-functions.log server-logs/edge-functions.log 2>/dev/null || true
cp -f $RUN_TMP/next-server.log server-logs/next-server.log 2>/dev/null || true
npx supabase stop --no-backup || true
- name: Upload server logs
# Use always() (not !cancelled()) so timeout-cancelled runs — exactly
# the case where these logs are most useful — still upload.
if: always()
uses: actions/upload-artifact@v4
with:
name: server-logs
path: server-logs/
retention-days: 14
- uses: actions/upload-artifact@v4
if: ${{ !cancelled() }}
with:
name: playwright-report
path: playwright-report/
retention-days: 30