harb/scripts/harb-evaluator/evaluate.sh
openhands 2ddd8e9ed2 fix: Holdout evaluator: Playwright browser-based scenario scripts (#381)
Replace shell-script scenario runner with Playwright. The evaluator now
runs `npx playwright test --config scripts/harb-evaluator/holdout.config.ts`
after booting the stack, using the existing tests/setup/ wallet-provider
and navigation infrastructure.

Changes:
- scripts/harb-evaluator/holdout.config.ts — new Playwright config pointing
  to scenarios/, headless chromium, 5-min timeout per test
- scripts/harb-evaluator/scenarios/sovereign-exit/always-leave.spec.ts —
  Playwright spec that buys KRK through the LocalSwapWidget then sells it
  back via the injected wallet provider, asserting sovereign exit works
- scripts/harb-evaluator/evaluate.sh — adds root npm install step (needed
  for npx playwright), exports STACK_* env aliases for getStackConfig(),
  replaces shell-script loop with a single playwright test invocation

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-03-01 11:24:15 +00:00

286 lines
12 KiB
Bash
Executable file

#!/usr/bin/env bash
# evaluate.sh — Spin up a fresh containerised harb stack from a PR branch,
# run holdout scenario scripts against it, then tear it down.
#
# Usage: evaluate.sh <pr-number>
#
# Exit codes:
# 0 gate passed (all scenarios succeeded, or no scenarios found)
# 1 gate failed (one or more scenario scripts returned non-zero)
# 2 infra error (stack failed to start, prerequisite missing, etc.)
#
# Environment overrides:
# HARB_REPO_REMOTE git remote to fetch from (default: origin)
# CODEBERG_REPO Gitea/Codeberg repo path (default: johba/harb)
#
# NOTE: host port isolation — docker-compose.yml binds fixed host ports
# (8545, 42069, 5173, 8081, 5100). Concurrent evaluation runs on the same
# host will collide on those ports. This script is designed for sequential use.
set -euo pipefail
# ── Constants ──────────────────────────────────────────────────────────
readonly REPO_REMOTE="${HARB_REPO_REMOTE:-origin}"
readonly CODEBERG_REPO="${CODEBERG_REPO:-johba/harb}"
readonly REPO_ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
readonly EVALUATOR_DIR="$(cd "$(dirname "$0")" && pwd)"
readonly SCENARIOS_DIR="$EVALUATOR_DIR/scenarios"
readonly ANVIL_TIMEOUT=120 # seconds to wait for anvil healthy
readonly BOOTSTRAP_TIMEOUT=180 # seconds to wait for bootstrap container exit
readonly PONDER_TIMEOUT=300 # seconds to wait for ponder /health
readonly PONDER_READY_TIMEOUT=360 # seconds to wait for ponder /ready (fully indexed)
readonly POLL_INTERVAL=5
# ── Logging helpers ────────────────────────────────────────────────────
log() { echo "[eval] $*"; }
infra_error() { echo "[eval] INFRA ERROR: $*" >&2; exit 2; }
gate_fail() { echo "[eval] GATE FAILED: $*" >&2; exit 1; }
usage() {
cat >&2 <<EOF
Usage: $0 <pr-number>
Checks out the PR branch into an isolated git worktree, boots a fresh
docker compose stack, runs scenario scripts, then tears down.
Exit codes:
0 gate passed
1 gate failed
2 infra error
EOF
exit 2
}
# ── Argument parsing ───────────────────────────────────────────────────
[[ $# -lt 1 ]] && usage
PR_NUMBER="$1"
[[ "$PR_NUMBER" =~ ^[0-9]+$ ]] || infra_error "Invalid PR number: '$PR_NUMBER'"
# ── Prerequisites ──────────────────────────────────────────────────────
if docker compose version &>/dev/null 2>&1; then
COMPOSE_CMD="docker compose"
# Compose v2 uses hyphens in container names: PROJECT-SERVICE-1
_COMPOSE_SEP="-"
elif command -v docker-compose &>/dev/null; then
COMPOSE_CMD="docker-compose"
# Compose v1 uses underscores in container names: PROJECT_SERVICE_1
_COMPOSE_SEP="_"
else
infra_error "docker compose not found. Install Docker with the compose plugin."
fi
command -v git &>/dev/null || infra_error "git not found"
command -v curl &>/dev/null || infra_error "curl not found"
# ── Fetch PR branch name ───────────────────────────────────────────────
# Try the Codeberg REST API first (requires ~/.netrc with credentials).
PR_BRANCH=""
_FETCHED_ALL=false
if [[ -f "$HOME/.netrc" ]]; then
log "Resolving PR #$PR_NUMBER branch via Codeberg API..."
api_json="$(curl --netrc --silent --max-time 10 \
"https://codeberg.org/api/v1/repos/$CODEBERG_REPO/pulls/$PR_NUMBER" 2>/dev/null)" || true
if [[ -n "$api_json" ]]; then
if command -v jq &>/dev/null; then
PR_BRANCH="$(echo "$api_json" | jq -r '.head.ref // empty' 2>/dev/null)" || true
else
# jq not available — use python3 for reliable nested key extraction.
# grep+sed is intentionally omitted: the Gitea response has multiple "ref"
# keys in nested objects and grep cannot safely target only head.ref.
PR_BRANCH="$(echo "$api_json" | \
python3 -c "import json,sys; print(json.load(sys.stdin)['head']['ref'])" 2>/dev/null)" || true
fi
fi
fi
# Fall back: fetch all remote refs and match common harb branch patterns.
if [[ -z "$PR_BRANCH" ]]; then
log "API lookup skipped or failed; scanning remote branches..."
cd "$REPO_ROOT"
git fetch "$REPO_REMOTE" --prune 2>/dev/null || infra_error "git fetch $REPO_REMOTE failed"
_FETCHED_ALL=true
PR_BRANCH="$(git branch -r 2>/dev/null | \
grep -E "(fix|feat|chore|refactor|hotfix)/.*[-/]${PR_NUMBER}[^0-9]?$|issue-${PR_NUMBER}$" | \
head -n1 | sed 's|.*/||' | tr -d ' ')" || true
fi
[[ -n "$PR_BRANCH" ]] || infra_error "Could not determine branch for PR #$PR_NUMBER"
log "PR #$PR_NUMBER => branch: $PR_BRANCH"
# ── Create isolated worktree ───────────────────────────────────────────
# Use mktemp -u to generate a unique path without creating the directory;
# git worktree add creates it. This avoids failures on older git that
# rejects a pre-existing (even empty) target directory.
WORKTREE_DIR="$(mktemp -u /tmp/harb-eval-${PR_NUMBER}-XXXXXX)"
# Use a project name that is unique per PR and safe for Docker labels.
COMPOSE_PROJECT="harb-eval-${PR_NUMBER}"
cleanup() {
local rc=$?
log "--- cleanup (exit $rc) ---"
if [[ -d "$WORKTREE_DIR" ]]; then
log "Tearing down stack (project: $COMPOSE_PROJECT)..."
(cd "$WORKTREE_DIR" && $COMPOSE_CMD -p "$COMPOSE_PROJECT" down -v --remove-orphans 2>/dev/null) || true
fi
log "Removing worktree $WORKTREE_DIR..."
cd "$REPO_ROOT"
git worktree remove --force "$WORKTREE_DIR" 2>/dev/null || rm -rf "$WORKTREE_DIR" || true
exit $rc
}
trap cleanup EXIT INT TERM
# Fetch the specific branch only if we haven't already fetched everything above.
cd "$REPO_ROOT"
if [[ "$_FETCHED_ALL" != "true" ]]; then
git fetch "$REPO_REMOTE" "$PR_BRANCH" 2>/dev/null || \
infra_error "Could not fetch branch '$PR_BRANCH' from $REPO_REMOTE"
fi
log "Creating worktree at $WORKTREE_DIR (branch: $PR_BRANCH)..."
git worktree add "$WORKTREE_DIR" "remotes/$REPO_REMOTE/$PR_BRANCH" \
|| infra_error "git worktree add failed for branch $PR_BRANCH"
# ── Build kraiken-lib in the worktree ─────────────────────────────────
log "Building kraiken-lib..."
(cd "$WORKTREE_DIR" && ./scripts/build-kraiken-lib.sh) \
|| infra_error "kraiken-lib build failed"
# ── Install root npm dependencies (needed for npx playwright test) ─────
log "Installing root npm dependencies..."
(cd "$WORKTREE_DIR" && npm install --no-audit --no-fund --silent) \
|| infra_error "npm install failed"
# ── Boot the stack ─────────────────────────────────────────────────────
cd "$WORKTREE_DIR"
log "Starting containerised stack (project: $COMPOSE_PROJECT)..."
$COMPOSE_CMD -p "$COMPOSE_PROJECT" up -d \
|| infra_error "docker compose up failed"
# Helper: resolve the container name for a service in this project.
# Compose v2 uses hyphens (PROJECT-SERVICE-1); v1 uses underscores (PROJECT_SERVICE_1).
container_name() { echo "${COMPOSE_PROJECT}${_COMPOSE_SEP}$1${_COMPOSE_SEP}1"; }
wait_healthy() {
local service="$1" timeout="$2"
local container
container="$(container_name "$service")"
log "Waiting for $service to be healthy (${timeout}s)..."
local deadline=$((SECONDS + timeout))
while (( SECONDS < deadline )); do
local status
status="$(docker inspect --format='{{.State.Health.Status}}' "$container" 2>/dev/null || echo "missing")"
if [[ "$status" == "healthy" ]]; then
log " $service healthy"
return 0
fi
sleep "$POLL_INTERVAL"
done
docker logs "$container" 2>&1 | tail -20 || true
infra_error "$service did not become healthy within ${timeout}s"
}
wait_exited() {
local service="$1" timeout="$2"
local container
container="$(container_name "$service")"
log "Waiting for $service container to complete (${timeout}s)..."
local deadline=$((SECONDS + timeout))
while (( SECONDS < deadline )); do
local status
status="$(docker inspect --format='{{.State.Status}}' "$container" 2>/dev/null || echo "missing")"
if [[ "$status" == "exited" || "$status" == "dead" ]]; then
local exit_code
exit_code="$(docker inspect --format='{{.State.ExitCode}}' "$container" 2>/dev/null || echo "1")"
if [[ "$exit_code" != "0" ]]; then
docker logs "$container" 2>&1 | tail -30 || true
infra_error "$service container exited with code $exit_code"
fi
log " $service completed successfully"
return 0
fi
sleep "$POLL_INTERVAL"
done
docker logs "$container" 2>&1 | tail -20 || true
infra_error "$service did not complete within ${timeout}s"
}
# Phase 1: base services
wait_healthy anvil "$ANVIL_TIMEOUT"
# Phase 2: bootstrap (deploys contracts, writes contracts.env)
wait_exited bootstrap "$BOOTSTRAP_TIMEOUT"
# ── Extract contract addresses ─────────────────────────────────────────
CONTRACTS_ENV="$WORKTREE_DIR/tmp/containers/contracts.env"
[[ -f "$CONTRACTS_ENV" ]] \
|| infra_error "contracts.env not found at $CONTRACTS_ENV"
log "Reading contract addresses from contracts.env..."
# shellcheck source=/dev/null
source "$CONTRACTS_ENV"
# Validate expected variables after sourcing — guards against set -u crashes
# if a future bootstrap refactor renames any of these.
KRAIKEN="${KRAIKEN:-}"
STAKE="${STAKE:-}"
LIQUIDITY_MANAGER="${LIQUIDITY_MANAGER:-}"
[[ -n "$KRAIKEN" ]] || infra_error "KRAIKEN not set in contracts.env"
[[ -n "$STAKE" ]] || infra_error "STAKE not set in contracts.env"
[[ -n "$LIQUIDITY_MANAGER" ]] || infra_error "LIQUIDITY_MANAGER not set in contracts.env"
log " KRAIKEN=$KRAIKEN"
log " STAKE=$STAKE"
log " LIQUIDITY_MANAGER=$LIQUIDITY_MANAGER"
# Phase 3: ponder must be healthy and fully indexed before running scenarios
wait_healthy ponder "$PONDER_TIMEOUT"
log "Waiting for Ponder to finish historical indexing (${PONDER_READY_TIMEOUT}s)..."
ponder_ready=false
ponder_deadline=$((SECONDS + PONDER_READY_TIMEOUT))
while (( SECONDS < ponder_deadline )); do
http_code="$(curl -sf -o /dev/null -w '%{http_code}' --max-time 3 \
http://127.0.0.1:42069/ready 2>/dev/null || echo "000")"
if [[ "$http_code" == "200" ]]; then
log " Ponder fully indexed"
ponder_ready=true
break
fi
sleep "$POLL_INTERVAL"
done
if [[ "$ponder_ready" != "true" ]]; then
infra_error "Ponder did not finish indexing within ${PONDER_READY_TIMEOUT}s"
fi
# ── Export stack endpoints ─────────────────────────────────────────────
export EVAL_PR_NUMBER="$PR_NUMBER"
export EVAL_BRANCH="$PR_BRANCH"
export EVAL_WORKTREE="$WORKTREE_DIR"
export EVAL_RPC_URL="http://127.0.0.1:8545"
export EVAL_GRAPHQL_URL="http://127.0.0.1:42069/graphql"
export EVAL_WEBAPP_URL="http://127.0.0.1:5173"
export EVAL_KRAIKEN="$KRAIKEN"
export EVAL_STAKE="$STAKE"
export EVAL_LIQUIDITY_MANAGER="$LIQUIDITY_MANAGER"
# Alias as STACK_* so getStackConfig() in tests/setup/stack.ts resolves correctly
export STACK_RPC_URL="$EVAL_RPC_URL"
export STACK_GRAPHQL_URL="$EVAL_GRAPHQL_URL"
export STACK_WEBAPP_URL="$EVAL_WEBAPP_URL"
log "Stack ready. Endpoints:"
log " RPC: $EVAL_RPC_URL"
log " GraphQL: $EVAL_GRAPHQL_URL"
log " WebApp: $EVAL_WEBAPP_URL"
# ── Run holdout Playwright scenarios ──────────────────────────────────
log "Running holdout scenarios via Playwright..."
cd "$WORKTREE_DIR"
if npx playwright test --config scripts/harb-evaluator/holdout.config.ts; then
log "Gate PASSED"
exit 0
else
gate_fail "One or more holdout scenarios failed"
fi