starwards-ci-debugging（starwards ci 调试）

CI Debugging for Starwards

Overview

Debug GitHub Actions CI failures independently by downloading logs and analyzing them locally.

Core principle: ALWAYS download CI logs before attempting to debug failures.

Prerequisites: Use the github-api skill to set up GitHub CLI access first.

The Iron Law

NO CI DEBUGGING WITHOUT LOG ACCESS FIRST

If you haven't downloaded logs, you cannot diagnose CI failures.

When to Use

Any Starwards GitHub Actions failure:

• Test-Static (TypeScript, ESLint, Prettier)
• Test-Units (Jest test failures)
• Test-E2e (Playwright test failures)
• Build failures
• Workflow configuration errors

Quick Start

Prerequisites:

1. Use github-api skill to install GitHub CLI
2. Verify authentication: gh auth status
3. Set alias: GH="/tmp/gh-install/gh_2.62.0_linux_amd64/bin/gh"

Quick Workflow:

# 1. Get PR checks
$GH pr checks <PR_NUMBER> --repo starwards/starwards

# 2. Extract run ID from URL
RUN_ID=$($GH pr checks <PR_NUMBER> --repo starwards/starwards | grep -oP 'runs/\K[0-9]+' | head -1)

# 3. Download logs
$GH api repos/starwards/starwards/actions/runs/$RUN_ID/logs --paginate > /tmp/run-logs.zip
cd /tmp && unzip -o -q run-logs.zip

# 4. Analyze failures (see sections below)

CI Log Analysis

Identify Failed Jobs

# Get failed jobs and steps
$GH api repos/starwards/starwards/actions/runs/<RUN_ID>/jobs \
  --jq '.jobs[] | select(.conclusion=="failure") | {name, steps: [.steps[] | select(.conclusion=="failure") | .name]}'

Example output:

{
  "name": "Test-Static",
  "steps": ["Run npm run test:format"]
}
{
  "name": "Test-E2e",
  "steps": ["Run npm run test:e2e"]
}

Log File Organization

Downloaded logs are named by job:

/tmp/0_Test-Units.txt
/tmp/1_Build.txt
/tmp/2_Test-E2e.txt
/tmp/3_Test-Static.txt

Test-Static Analysis

# Find ESLint errors
grep "##\[error\]" /tmp/3_Test-Static.txt

# Find TypeScript errors
grep "error TS" /tmp/3_Test-Static.txt

# Find Prettier failures
grep -A 5 "Code style issues" /tmp/3_Test-Static.txt

# Group TypeScript errors by code
grep "error TS" /tmp/3_Test-Static.txt | grep -oP 'error TS\d+' | sort | uniq -c

Test-E2e Analysis

# Find Playwright errors
grep -E "FAIL|Error:" /tmp/2_Test-E2e.txt | head -50

# Find browser launch errors
grep "Executable doesn't exist" /tmp/2_Test-E2e.txt

# Find test assertion failures
grep -A 10 "Expected.*Received" /tmp/2_Test-E2e.txt

# Find timeout issues
grep -i "timeout" /tmp/2_Test-E2e.txt

Test-Units Analysis

# Find test failures
grep -A 20 "FAIL " /tmp/0_Test-Units.txt

# Find timeout errors
grep "Timeout" /tmp/0_Test-Units.txt

# Find assertion errors
grep -B 5 -A 10 "expect.*toEqual\|toBe\|toBeCloseTo" /tmp/0_Test-Units.txt

# Find connection errors (common in Colyseus tests)
grep -i "connection.*error\|ECONNREFUSED" /tmp/0_Test-Units.txt

Common Starwards CI Failures

1. ESLint: Unsafe any Type Usage

Symptoms:

##[error]  192:23  error  Unsafe assignment of an `any` value  @typescript-eslint/no-unsafe-assignment
##[error]  192:61  error  Unexpected any. Specify a different type  @typescript-eslint/no-explicit-any

Root cause: Using (e as any).code violates strict TypeScript rules.

Fix pattern:

// BEFORE (fails lint):
const hasCode = typeof (e as any).code === 'string' && (e as any).code;
const code = String((e as any).code);

// AFTER (passes lint):
function hasStringCode(e: unknown): e is { code: string } {
    return typeof (e as { code?: unknown })?.code === 'string'
        && !!(e as { code?: string }).code;
}

if (hasStringCode(e)) {
    const code = e.code; // Type-safe!
}

Local verification:

npm run lint
# Expected: No errors

2. Playwright: Browser Executable Not Found

Symptoms:

Error: browserType.launch: Executable doesn't exist at /ms-playwright/chromium_headless_shell-1194/chrome-linux/headless_shell

Root cause: Docker image version doesn't match installed Playwright version.

Diagnosis:

# Check package version
grep '"@playwright/test"' package.json
# Example: "@playwright/test": "^1.56.1"

# Check exact installed version
grep -A 15 'node_modules/@playwright/test"' package-lock.json | grep '"version"' | head -1
# Example: "version": "1.56.1"

# Check CI workflow Docker image
grep "mcr.microsoft.com/playwright" .github/workflows/ci-cd.yml
# Example: image: mcr.microsoft.com/playwright:v1.42.1-jammy
#                                                 ^^^^^^^ MISMATCH!

Fix:

# .github/workflows/ci-cd.yml
Test-E2e:
    runs-on: ubuntu-latest
    container:
        image: mcr.microsoft.com/playwright:v1.56.1-jammy  # Match package version exactly
        options: --user 1001

Find available versions:

curl -s https://mcr.microsoft.com/v2/playwright/tags/list | jq -r '.tags[] | select(startswith("v1.5"))' | sort -V | tail -10

Local verification:

# Cannot test Docker image locally, must verify on CI
git add .github/workflows/ci-cd.yml
git commit -m "Update Playwright Docker image to v1.56.1"
git push

3. TypeScript: Type Errors

Symptoms:

modules/core/src/file.ts(42,15): error TS2322: Type 'string | undefined' is not assignable to type 'string'

Common error codes:

• TS2322: Type mismatch - add type assertion or fix types
• TS2532: Possibly undefined - add null check or ! assertion
• TS2345: Argument type mismatch - check function signature
• TS2339: Property doesn't exist - check interface/type definition

Local verification:

npm run test:types
# Expected: No errors

4. Jest: Flaky Colyseus Tests

Symptoms:

connection error: AggregateError
expect(received).toEqual(expected)
  "text": "err: "  // Expected "err: ECONNREFUSED"

Common causes:

• Dependency update changed error message format
• Race condition in async test
• Port conflict (rare)
• Timing issue in Colyseus state sync

Diagnosis:

# Check recent changes
git log --oneline -5
git diff HEAD~1 package.json

# Search for error in logs
grep -B 10 -A 10 "expect.*toEqual" /tmp/0_Test-Units.txt

# Look for timing issues
grep -i "timeout\|race\|async" /tmp/0_Test-Units.txt

Fix pattern:

• Update expected error format to match new behavior
• Add longer timeouts for flaky tests
• Use waitFor helpers from @testing-library/react
• Check Colyseus state sync timing

5. Build Failures

Symptoms:

error TS6059: File '...' is not under 'rootDir'
error TS5055: Cannot write file '...' because it would overwrite input file

Common causes:

• Circular dependencies
• Module import issues
• Monorepo build order violation
• Missing npm run build:core before dependent modules

Diagnosis:

# Check build order
grep "npm run build" .github/workflows/ci-cd.yml

# Check for circular deps
npm ls | grep "deduped"

# Verify module dependencies
cat modules/browser/package.json | jq '.dependencies'

Fix:

• Use starwards-monorepo skill for build order issues
• Check imports in affected files
• Ensure core builds before browser/server

Complete CI Debug Workflow

Full Script

#!/bin/bash
set -e

# Setup (use github-api skill for this part)
GH="/tmp/gh-install/gh_2.62.0_linux_amd64/bin/gh"
REPO="starwards/starwards"
PR_NUMBER=$1

if [ -z "$PR_NUMBER" ]; then
  echo "Usage: $0 <PR_NUMBER>"
  exit 1
fi

echo "=== Step 1: Check PR Status ==="
$GH pr view $PR_NUMBER --repo $REPO

echo -e "\n=== Step 2: Get Check Results ==="
$GH pr checks $PR_NUMBER --repo $REPO

echo -e "\n=== Step 3: Identify Failed Jobs ==="
RUN_ID=$($GH pr checks $PR_NUMBER --repo $REPO | grep -oP 'runs/\K[0-9]+' | head -1)
echo "Run ID: $RUN_ID"

$GH api repos/$REPO/actions/runs/$RUN_ID/jobs \
  --jq '.jobs[] | select(.conclusion=="failure") | {
    name: .name,
    failed_steps: [.steps[] | select(.conclusion=="failure") | .name]
  }'

echo -e "\n=== Step 4: Download Logs ==="
$GH api repos/$REPO/actions/runs/$RUN_ID/logs --paginate > /tmp/run-logs.zip
cd /tmp && unzip -o -q run-logs.zip
echo "Logs extracted to /tmp/*.txt"

echo -e "\n=== Step 5: Analyze Failures ==="
echo "TypeScript errors:"
grep -h "error TS" /tmp/*.txt | head -10 || echo "None found"

echo -e "\nESLint errors:"
grep -h "##\[error\]" /tmp/*.txt | head -10 || echo "None found"

echo -e "\nTest failures:"
grep -h "FAIL.*spec\." /tmp/*.txt | head -10 || echo "None found"

echo -e "\nBrowser errors:"
grep -h "Executable doesn't exist" /tmp/*.txt || echo "None found"

echo -e "\n=== Step 6: Next Steps ==="
echo "1. Review errors above"
echo "2. Fix issues locally"
echo "3. Run: npm run lint && npm test && npm run build"
echo "4. Commit and push"
echo "5. Monitor new CI run with: $GH pr checks $PR_NUMBER --repo $REPO"

Save and Use

# Save script
cat > /tmp/debug-ci.sh << 'EOF'
[paste script above]
EOF

chmod +x /tmp/debug-ci.sh

# Run
/tmp/debug-ci.sh 1826

Local Verification Checklist

Before pushing CI fixes:

# ☐ Lint passes
npm run lint

# ☐ Type check passes
npm run test:types

# ☐ Unit tests pass
npm test

# ☐ Build succeeds
npm run build

# ☐ E2E tests pass (if applicable)
npm run test:e2e

# ☐ Format is correct
npm run test:format

Monitoring After Push

# Wait for CI to start
sleep 30

# Check new run status
$GH pr checks <PR_NUMBER> --repo starwards/starwards

# If still failing, download new logs and repeat
RUN_ID=$($GH pr checks <PR_NUMBER> --repo starwards/starwards | grep -oP 'runs/\K[0-9]+' | head -1)
$GH api repos/starwards/starwards/actions/runs/$RUN_ID/logs --paginate > /tmp/run-logs-retry.zip

Integration with Other Skills

• github-api - Setup GitHub CLI, download logs, query API
• starwards-debugging - Systematic debugging for local reproduction
• starwards-verification - Run verification commands before push
• starwards-monorepo - Understand build dependencies and order
• starwards-workflow - Terminal commands for local testing

Quick Reference

GitHub CLI Commands

See github-api skill for:

• Installation and setup
• Authentication
• API access patterns
• Rate limiting

CI-Specific Commands

# Get PR checks
$GH pr checks <PR> --repo starwards/starwards

# Get run details
$GH run view <RUN_ID> --repo starwards/starwards

# Download logs
$GH api repos/starwards/starwards/actions/runs/<RUN_ID>/logs --paginate > /tmp/logs.zip

# Get failed jobs
$GH api repos/starwards/starwards/actions/runs/<RUN_ID>/jobs \
  --jq '.jobs[] | select(.conclusion=="failure")'

Log Analysis Patterns

# ESLint errors
grep "##\[error\]" /tmp/*.txt

# TypeScript errors
grep "error TS" /tmp/*.txt

# Test failures
grep "FAIL " /tmp/*.txt

# Playwright errors
grep "Executable doesn't exist" /tmp/*.txt

# Assertion failures
grep -A 10 "Expected.*Received" /tmp/*.txt

Tips

1. Download logs first - Always get complete context before debugging
2. Check run ID - Ensure you're analyzing the correct workflow run
3. Group errors - Use sort | uniq -c to identify patterns
4. Verify locally - Run all checks before pushing
5. Monitor after push - Watch new CI run to confirm fix
6. Check dependencies - Build order matters in monorepos
7. Save scripts - Keep debug workflow script for repeated use

Real-World Example

Scenario: PR #1826 failing with Test-Static and Test-E2e

# Setup (github-api skill)
GH="/tmp/gh-install/gh_2.62.0_linux_amd64/bin/gh"

# Check status
$GH pr checks 1826 --repo starwards/starwards
# Test-Static: FAIL
# Test-E2e: FAIL

# Download logs
RUN_ID=19084760492
$GH api repos/starwards/starwards/actions/runs/$RUN_ID/logs --paginate > /tmp/run-logs.zip
cd /tmp && unzip -o -q run-logs.zip

# Analyze Test-Static
grep "##\[error\]" /tmp/3_Test-Static.txt
# Found: 7 ESLint errors about unsafe `any` usage
# File: modules/core/src/client/connection-manager.ts line 192

# Analyze Test-E2e
grep "Executable doesn't exist" /tmp/2_Test-E2e.txt
# Found: Playwright browser missing
# Cause: Docker image v1.42.1 vs package v1.56.1

# Fix: Add type guard for ESLint
# Fix: Update .github/workflows/ci-cd.yml to use playwright:v1.56.1-jammy

# Verify locally
npm run lint      # ✓ Passes
npm run test:types # ✓ Passes

# Push
git add -A
git commit -m "Fix CI failures: eslint errors and Playwright version mismatch"
git push

# Monitor
sleep 30
$GH pr checks 1826 --repo starwards/starwards
# Result: All CI checks pass ✅