Files
ComfyUI_frontend/scripts/cicd/pr-playwright-deploy-and-comment.sh
snomiao 48d01745cd feat: add test count display to Playwright PR comments (#5458)
* feat: add test count display to Playwright PR comments

- Add extract-playwright-counts.mjs script to parse test results from Playwright reports
- Update pr-playwright-deploy-and-comment.sh to extract and display test counts
- Show overall summary with passed/failed/flaky/skipped counts
- Display per-browser test counts inline with report links
- Use dynamic status icons based on test results (//⚠️)

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* feat: include skipped test count in per-browser display

- Add skipped test extraction for individual browser reports
- Update per-browser display format to show all four counts:
  ( passed /  failed / ⚠️ flaky / ⏭️ skipped)
- Provides complete test result visibility at a glance

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* fix: improve test count extraction reliability in CI

- Use absolute paths for script and report directories
- Add debug logging to help diagnose extraction issues
- Move counts display after View Report link as requested
- Format: [View Report](url) •  passed /  failed / ⚠️ flaky / ⏭️ skipped

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* fix: generate JSON reports alongside HTML for test count extraction

- Add JSON reporter to Playwright test runs
- Generate report.json alongside HTML reports
- Store JSON report in playwright-report directory
- This enables accurate test count extraction from CI artifacts

The HTML reports alone don't contain easily extractable test statistics
as they use a React app with dynamically loaded data. JSON reports
provide direct access to test counts.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* fix: correct JSON reporter syntax for Playwright tests

- Use proper syntax for JSON reporter with outputFile option
- Run separate commands for HTML and JSON report merging
- Specify output path directly in reporter configuration
- Ensures report.json is created in playwright-report directory

This fixes the "No such file or directory" error when trying to move
report.json file, as it wasn't being created in the first place.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Revert "fix: correct JSON reporter syntax for Playwright tests"

This reverts commit 605d7cc1e2.

* fix: use correct Playwright reporter syntax with comma-separated list

- Use --reporter=html,json syntax (comma-separated, not space)
- Move test-results.json to playwright-report/report.json after generation
- Remove incorrect PLAYWRIGHT_JSON_OUTPUT_NAME env variable
- Add || true to prevent failure if JSON file doesn't exist

The JSON reporter outputs to test-results.json by default when using
the comma-separated reporter list syntax.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* fix: improve test count extraction reliability in CI

- Use separate --reporter flags for list, html, and json
- Set PLAYWRIGHT_JSON_OUTPUT_NAME env var to specify JSON output path
- Run HTML and JSON report generation separately for merged reports
- Ensures report.json is created in playwright-report directory

The combined reporter syntax wasn't creating the JSON file properly.
Using separate reporter flags with env var ensures JSON is generated.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* Update scripts/cicd/pr-playwright-deploy-and-comment.sh

Co-authored-by: Alexander Brown <drjkl@comfy.org>

* refactor: convert extraction script to TypeScript and use tsx

- Convert extract-playwright-counts.mjs to TypeScript (.ts)
- Add proper TypeScript types for better type safety
- Use tsx for execution instead of node
- Auto-install tsx in CI if not available
- Better alignment with the TypeScript codebase

This provides better type safety and consistency with the rest of
the codebase while maintaining the same functionality.

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

* chore(pr-playwright-deploy-and-comment.sh): move tsx installation check to the beginning of the script for better organization and efficiency

* [auto-fix] Apply ESLint and Prettier fixes

---------

Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Alexander Brown <drjkl@comfy.org>
Co-authored-by: GitHub Action <action@github.com>
2025-09-13 01:18:18 -07:00

377 lines
13 KiB
Bash
Executable File

#!/bin/bash
set -e
# Deploy Playwright test reports to Cloudflare Pages and comment on PR
# Usage: ./pr-playwright-deploy-and-comment.sh <pr_number> <branch_name> <status> [start_time]
# Input validation
# Validate PR number is numeric
case "$1" in
''|*[!0-9]*)
echo "Error: PR_NUMBER must be numeric" >&2
exit 1
;;
esac
PR_NUMBER="$1"
# Sanitize and validate branch name (allow alphanumeric, dots, dashes, underscores, slashes)
BRANCH_NAME=$(echo "$2" | sed 's/[^a-zA-Z0-9._/-]//g')
if [ -z "$BRANCH_NAME" ]; then
echo "Error: Invalid or empty branch name" >&2
exit 1
fi
# Validate status parameter
STATUS="${3:-completed}"
case "$STATUS" in
starting|completed) ;;
*)
echo "Error: STATUS must be 'starting' or 'completed'" >&2
exit 1
;;
esac
START_TIME="${4:-$(date -u '+%m/%d/%Y, %I:%M:%S %p')}"
# Required environment variables
: "${GITHUB_TOKEN:?GITHUB_TOKEN is required}"
: "${GITHUB_REPOSITORY:?GITHUB_REPOSITORY is required}"
# Cloudflare variables only required for deployment
if [ "$STATUS" = "completed" ]; then
: "${CLOUDFLARE_API_TOKEN:?CLOUDFLARE_API_TOKEN is required for deployment}"
: "${CLOUDFLARE_ACCOUNT_ID:?CLOUDFLARE_ACCOUNT_ID is required for deployment}"
fi
# Configuration
COMMENT_MARKER="<!-- PLAYWRIGHT_TEST_STATUS -->"
# Use dot notation for artifact names (as Playwright creates them)
BROWSERS="chromium chromium-2x chromium-0.5x mobile-chrome"
# Install wrangler if not available (output to stderr for debugging)
if ! command -v wrangler > /dev/null 2>&1; then
echo "Installing wrangler v4..." >&2
npm install -g wrangler@^4.0.0 >&2 || {
echo "Failed to install wrangler" >&2
echo "failed"
return
}
fi
# Check if tsx is available, install if not
if ! command -v tsx > /dev/null 2>&1; then
echo "Installing tsx..." >&2
npm install -g tsx >&2 || echo "Failed to install tsx" >&2
fi
# Deploy a single browser report, WARN: ensure inputs are sanitized before calling this function
deploy_report() {
dir="$1"
browser="$2"
branch="$3"
[ ! -d "$dir" ] && echo "failed" && return
# Project name with dots converted to dashes for Cloudflare
sanitized_browser=$(echo "$browser" | sed 's/\./-/g')
project="comfyui-playwright-${sanitized_browser}"
echo "Deploying $browser to project $project on branch $branch..." >&2
# Try deployment up to 3 times
i=1
while [ $i -le 3 ]; do
echo "Deployment attempt $i of 3..." >&2
# Branch and project are already sanitized, use them directly
# Branch was sanitized at script start, project uses sanitized_browser
if output=$(wrangler pages deploy "$dir" \
--project-name="$project" \
--branch="$branch" 2>&1); then
# Extract URL from output (improved regex for valid URL characters)
url=$(echo "$output" | grep -oE 'https://[a-zA-Z0-9.-]+\.pages\.dev\S*' | head -1)
result="${url:-https://${branch}.${project}.pages.dev}"
echo "Success! URL: $result" >&2
echo "$result" # Only this goes to stdout for capture
return
else
echo "Deployment failed on attempt $i: $output" >&2
fi
[ $i -lt 3 ] && sleep 10
i=$((i + 1))
done
echo "failed"
}
# Post or update GitHub comment
post_comment() {
body="$1"
temp_file=$(mktemp)
echo "$body" > "$temp_file"
if command -v gh > /dev/null 2>&1; then
# Find existing comment ID
existing=$(gh api "repos/$GITHUB_REPOSITORY/issues/$PR_NUMBER/comments" \
--jq ".[] | select(.body | contains(\"$COMMENT_MARKER\")) | .id" | head -1)
if [ -n "$existing" ]; then
# Update specific comment by ID
gh api --method PATCH "repos/$GITHUB_REPOSITORY/issues/comments/$existing" \
--field body="$(cat "$temp_file")"
else
# Create new comment
gh pr comment "$PR_NUMBER" --body-file "$temp_file"
fi
else
echo "GitHub CLI not available, outputting comment:"
cat "$temp_file"
fi
rm -f "$temp_file"
}
# Main execution
if [ "$STATUS" = "starting" ]; then
# Post starting comment
comment=$(cat <<EOF
$COMMENT_MARKER
## 🎭 Playwright Test Results
<img alt='loading' src='https://github.com/user-attachments/assets/755c86ee-e445-4ea8-bc2c-cca85df48686' width='14px' height='14px'/> **Tests are starting...**
⏰ Started at: $START_TIME UTC
### 🚀 Running Tests
- 🧪 **chromium**: Running tests...
- 🧪 **chromium-0.5x**: Running tests...
- 🧪 **chromium-2x**: Running tests...
- 🧪 **mobile-chrome**: Running tests...
---
⏱️ Please wait while tests are running...
EOF
)
post_comment "$comment"
else
# Deploy and post completion comment
# Convert branch name to Cloudflare-compatible format (lowercase, only alphanumeric and dashes)
cloudflare_branch=$(echo "$BRANCH_NAME" | tr '[:upper:]' '[:lower:]' | \
sed 's/[^a-z0-9-]/-/g' | sed 's/--*/-/g' | sed 's/^-\|-$//g')
echo "Looking for reports in: $(pwd)/reports"
echo "Available reports:"
ls -la reports/ 2>/dev/null || echo "Reports directory not found"
# Deploy all reports in parallel and collect URLs + test counts
temp_dir=$(mktemp -d)
pids=""
i=0
# Store current working directory for absolute paths
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
BASE_DIR="$(pwd)"
# Start parallel deployments and count extractions
for browser in $BROWSERS; do
if [ -d "reports/playwright-report-$browser" ]; then
echo "Found report for $browser, deploying in parallel..."
(
url=$(deploy_report "reports/playwright-report-$browser" "$browser" "$cloudflare_branch")
echo "$url" > "$temp_dir/$i.url"
echo "Deployment result for $browser: $url"
# Extract test counts using tsx (TypeScript executor)
EXTRACT_SCRIPT="$SCRIPT_DIR/extract-playwright-counts.ts"
REPORT_DIR="$BASE_DIR/reports/playwright-report-$browser"
if command -v tsx > /dev/null 2>&1 && [ -f "$EXTRACT_SCRIPT" ]; then
echo "Extracting counts from $REPORT_DIR using $EXTRACT_SCRIPT" >&2
counts=$(tsx "$EXTRACT_SCRIPT" "$REPORT_DIR" 2>&1 || echo '{}')
echo "Extracted counts for $browser: $counts" >&2
echo "$counts" > "$temp_dir/$i.counts"
else
echo "Script not found or tsx not available: $EXTRACT_SCRIPT" >&2
echo '{}' > "$temp_dir/$i.counts"
fi
) &
pids="$pids $!"
else
echo "Report not found for $browser at reports/playwright-report-$browser"
echo "failed" > "$temp_dir/$i.url"
echo '{}' > "$temp_dir/$i.counts"
fi
i=$((i + 1))
done
# Wait for all deployments to complete
for pid in $pids; do
wait $pid
done
# Collect URLs and counts in order
urls=""
all_counts=""
i=0
for browser in $BROWSERS; do
if [ -f "$temp_dir/$i.url" ]; then
url=$(cat "$temp_dir/$i.url")
else
url="failed"
fi
if [ -z "$urls" ]; then
urls="$url"
else
urls="$urls $url"
fi
if [ -f "$temp_dir/$i.counts" ]; then
counts=$(cat "$temp_dir/$i.counts")
echo "Read counts for $browser from $temp_dir/$i.counts: $counts" >&2
else
counts="{}"
echo "No counts file found for $browser at $temp_dir/$i.counts" >&2
fi
if [ -z "$all_counts" ]; then
all_counts="$counts"
else
all_counts="$all_counts|$counts"
fi
i=$((i + 1))
done
# Clean up temp directory
rm -rf "$temp_dir"
# Calculate total test counts across all browsers
total_passed=0
total_failed=0
total_flaky=0
total_skipped=0
total_tests=0
# Parse counts and calculate totals
IFS='|'
set -- $all_counts
for counts_json; do
if [ "$counts_json" != "{}" ] && [ -n "$counts_json" ]; then
# Parse JSON counts using simple grep/sed if jq is not available
if command -v jq > /dev/null 2>&1; then
passed=$(echo "$counts_json" | jq -r '.passed // 0')
failed=$(echo "$counts_json" | jq -r '.failed // 0')
flaky=$(echo "$counts_json" | jq -r '.flaky // 0')
skipped=$(echo "$counts_json" | jq -r '.skipped // 0')
total=$(echo "$counts_json" | jq -r '.total // 0')
else
# Fallback parsing without jq
passed=$(echo "$counts_json" | sed -n 's/.*"passed":\([0-9]*\).*/\1/p')
failed=$(echo "$counts_json" | sed -n 's/.*"failed":\([0-9]*\).*/\1/p')
flaky=$(echo "$counts_json" | sed -n 's/.*"flaky":\([0-9]*\).*/\1/p')
skipped=$(echo "$counts_json" | sed -n 's/.*"skipped":\([0-9]*\).*/\1/p')
total=$(echo "$counts_json" | sed -n 's/.*"total":\([0-9]*\).*/\1/p')
fi
total_passed=$((total_passed + ${passed:-0}))
total_failed=$((total_failed + ${failed:-0}))
total_flaky=$((total_flaky + ${flaky:-0}))
total_skipped=$((total_skipped + ${skipped:-0}))
total_tests=$((total_tests + ${total:-0}))
fi
done
unset IFS
# Determine overall status
if [ $total_failed -gt 0 ]; then
status_icon="❌"
status_text="Some tests failed"
elif [ $total_flaky -gt 0 ]; then
status_icon="⚠️"
status_text="Tests passed with flaky tests"
elif [ $total_tests -gt 0 ]; then
status_icon="✅"
status_text="All tests passed!"
else
status_icon="🕵🏻"
status_text="No test results found"
fi
# Generate completion comment
comment="$COMMENT_MARKER
## 🎭 Playwright Test Results
$status_icon **$status_text**
⏰ Completed at: $(date -u '+%m/%d/%Y, %I:%M:%S %p') UTC"
# Add summary counts if we have test data
if [ $total_tests -gt 0 ]; then
comment="$comment
### 📈 Summary
- **Total Tests:** $total_tests
- **Passed:** $total_passed
- **Failed:** $total_failed $([ $total_failed -gt 0 ] && echo '❌' || echo '')
- **Flaky:** $total_flaky $([ $total_flaky -gt 0 ] && echo '⚠️' || echo '')
- **Skipped:** $total_skipped $([ $total_skipped -gt 0 ] && echo '⏭️' || echo '')"
fi
comment="$comment
### 📊 Test Reports by Browser"
# Add browser results with individual counts
i=0
IFS='|'
set -- $all_counts
for counts_json; do
# Get browser name
browser=$(echo "$BROWSERS" | cut -d' ' -f$((i + 1)))
# Get URL at position i
url=$(echo "$urls" | cut -d' ' -f$((i + 1)))
if [ "$url" != "failed" ] && [ -n "$url" ]; then
# Parse individual browser counts
if [ "$counts_json" != "{}" ] && [ -n "$counts_json" ]; then
if command -v jq > /dev/null 2>&1; then
b_passed=$(echo "$counts_json" | jq -r '.passed // 0')
b_failed=$(echo "$counts_json" | jq -r '.failed // 0')
b_flaky=$(echo "$counts_json" | jq -r '.flaky // 0')
b_skipped=$(echo "$counts_json" | jq -r '.skipped // 0')
b_total=$(echo "$counts_json" | jq -r '.total // 0')
else
b_passed=$(echo "$counts_json" | sed -n 's/.*"passed":\([0-9]*\).*/\1/p')
b_failed=$(echo "$counts_json" | sed -n 's/.*"failed":\([0-9]*\).*/\1/p')
b_flaky=$(echo "$counts_json" | sed -n 's/.*"flaky":\([0-9]*\).*/\1/p')
b_skipped=$(echo "$counts_json" | sed -n 's/.*"skipped":\([0-9]*\).*/\1/p')
b_total=$(echo "$counts_json" | sed -n 's/.*"total":\([0-9]*\).*/\1/p')
fi
if [ -n "$b_total" ] && [ "$b_total" != "0" ]; then
counts_str=" • ✅ $b_passed / ❌ $b_failed / ⚠️ $b_flaky / ⏭️ $b_skipped"
else
counts_str=""
fi
else
counts_str=""
fi
comment="$comment
- ✅ **${browser}**: [View Report](${url})${counts_str}"
else
comment="$comment
- ❌ **${browser}**: Deployment failed"
fi
i=$((i + 1))
done
unset IFS
comment="$comment
---
🎉 Click on the links above to view detailed test results for each browser configuration."
post_comment "$comment"
fi