makefile(web): auto-heal Playwright install and preserve output for s…

[rodneyaquino]

🧩 PR: Improve Playwright Web Test Reliability and Visibility

🎯 Summary

This update enhances the test-web target in the Makefile to make Playwright runs more self-healing and transparent. If the local CLI or dependencies are missing, it now auto-installs them and continues the test flow.

✅ What Changed

Makefile – test-web target

test-web: ## Run Playwright E2E (includes a11y)
	@printf "$(BLUE)🌐 Web E2E (Playwright)$(NC)\n"
	@cd tests/web && \
		if [ ! -x node_modules/.bin/playwright ]; then \
			printf "$(YELLOW)⚠️  Playwright not found. Installing deps...$(NC)\n"; \
			npm install && npx playwright install --with-deps || npx playwright install; \
		fi
	@cd tests/web && npm test 2>&1 | tee $(abspath $(WEB_LOG)) || { printf "$(RED)❌ Web failed$(NC)\n"; exit 1; }
	@perl -i -pe 's/\e\[[0-9;]*[A-Za-z]//g' $(WEB_LOG) || true
	@printf "$(GREEN)✅ Web passed$(NC)\n"

💡 Why

• Before: Playwright CLI wasn't always available in fresh clones or CI runs. Missing binaries caused sh: playwright: command not found errors and empty logs.
• Now: The task self-checks for node_modules/.bin/playwright, installs deps if absent, and cleans ANSI codes for reliable log parsing.
• Result:
  • No more manual npm install before running make test-all
  • Full Playwright output visible in console and web-results.txt
  • Skipped test counts now display cleanly in summary

🧠 Notes

• Keeps Makefile behavior consistent across local and CI runs
• Continues to log everything to test-results/web-results.txt
• Safe re-entry: repeated runs skip reinstall if already present

🧪 Example Output

🌐 Web E2E (Playwright)
⚠️  Playwright not found. Installing deps...
Running 7 tests using 4 workers
  ✓ WEB-001: login with valid credentials shows inventory page (1.9s)
  - WEB-A11Y-KNOWN-BUG-001: header sort dropdown is missing an accessible name
  ✓ WEB-004: sorting A→Z then Z→A changes first visible item (1.3s)
  1 skipped
  6 passed (4.4s)

✅ Web passed

🔍 Verification

• Ran make test-web locally on a clean machine
• Verified skipped test counts appear in make summary
• Confirmed exit codes propagate correctly (fail on real test errors)