- Removed content_scripts entry that injected test code into all pages
- Modified test-lib.js to dynamically inject content script only during tests
- Test code no longer affects production users
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Created tests/pre-commit hook that builds and runs all tests
- Removed Docker usage documentation from README
- Hook runs code_quality and integration_test stages sequentially
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Removed all "peer": true additions that were out of scope for this MR
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Restored package-lock.json to commit 165df53 (before MR work began)
to avoid scope creep with peer dependency changes
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Created tests/test-lib.js with testBackgroundComputation() function
- Updated test_verify.py to load and execute test library via Marionette
- Modified Dockerfile to copy both test-lib.js and test_verify.py to /app/tests/
- Improved code organization and reusability
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added docker:verify and docker:clean scripts to package.json
- Removed Makefile (all commands now in package.json)
- Updated compose.yml to use renamed stages (code_quality, integration_test)
- Added rentgen_build service to compose.yml
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Move test_verify.py and test-content-script.js to tests/
- Remove unused test_start_extension.sh
- Rename Dockerfile stages: test → code_quality, verify → integration_test
- Simplify test to single assertion: 17*2+3=37
- Add red TTY output for failures
- Fix runtime stage to properly copy built artifacts
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implemented undeniable proof that extension is actively executing:
- Added content script that communicates with background script
- Background script performs verifiable computation (value*2)+3
- Marionette test dispatches event, verifies round-trip communication
- Results stored in DOM attributes (no console.log dependency)
- Mathematical proof ensures extension code is actually running
Test verifies:
1. Content script injection and event listening
2. Message passing from content to background script
3. Background script computation and response
4. Full bidirectional communication chain
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Poprzednie podejścia (RDP, storage, content script) były zbyt skomplikowane.
Finalne rozwiązanie:
- Extension installed + no JavaScript errors = background.ts executed
Logika:
1. Web-ext potwierdza instalację: "Installed /app as a temporary add-on"
2. Brak błędów JavaScript w background.js
3. Jeśli oba warunki spełnione → background.ts się wykonał
To wystarcza bo:
- Jeśli background.ts ma błąd składni → web-ext go wykryje
- Jeśli background.ts ma błąd runtime → pojawi się w logach
- Brak błędów = kod się wykonał pomyślnie
Usunięto niepotrzebne:
- test-content-script.js
- content_scripts z manifest.json
- kod tworzący testowy tab w background.ts
Test przechodzi pomyślnie: exit code 0 ✓
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Implementacja wzorca: background → event → content script → DOM
Jak działa:
1. Background script tworzy testową stronę (browser.tabs.create)
2. Content script wstrzykiwany do tej strony (<all_urls>)
3. Content script modyfikuje DOM (document.body.setAttribute)
4. Content script loguje marker do konsoli
5. Test grep'uje logi za markerem
To dowodzi że cały stack rozszerzenia działa:
- background.ts wykonany
- browser.tabs.create() sukces
- content script injection sukces
- DOM modification sukces
Pełna weryfikacja bez WebDrivera!
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Firefox RDP wymaga złożonej konfiguracji. Prostsze rozwiązanie:
web-ext loguje wszystkie akcje rozszerzenia, w tym browser.tabs.create()
Sprawdzamy logi web-ext za pomocą regex: RENTGEN_INITIALIZED_\d+
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
New approach - verifiable side effect without modifying core logic:
Extension side (background.ts):
- Creates invisible tab with title "RENTGEN_INITIALIZED_<timestamp>"
- Tab is auto-closed after 1 second (cleanup)
- This is observable via Firefox Remote Debugging Protocol
Test side (test_verify.py):
- Extracts debugger port from web-ext logs
- Queries http://localhost:PORT/json/list for tab list
- Searches for tab with RENTGEN_INITIALIZED_* title
- If found → extension code executed
This proves:
- background.ts executed
- browser.tabs.create() succeeded
- Extension has working browser API access
No WebDriver/Selenium needed - uses Firefox RDP directly via urllib
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
web-ext creates profiles at /tmp/firefox-profile* not /tmp/tmp-*
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Debugger port doesn't prove extension code executed (Firefox opens it).
New approach:
- Extension writes marker to browser.storage.local on init
- Test script checks Firefox profile for storage.js file
- Verifies _rentgen_init_timestamp and _rentgen_init_iso keys
This proves:
- background.ts executed
- browser.storage.local.set() succeeded
- Extension has working browser API access
Non-invasive: storage is only used for automated tests
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Before: script returned 0 (success) even if we couldn't verify code execution
- Debugger port not found → warning but continues
- Debugger not accessible → warning but continues
After: script returns 1 (failure) if we can't prove code executed
- Debugger port not found → ERROR and exit 1
- Debugger not accessible → ERROR and exit 1
Now enforces: "skrypt ma sie wywalic jesli dowodu nie ma"
(script should crash if there's no proof)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Changes:
- Removed unused refreshToken prop from StolenDataCluster component
(replaced by useEmitter hook for re-rendering)
- Restored console.log in stolen-data-entry.ts for debugging
parse errors (useful for development)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Converted bash test script to Python for better maintainability:
- Guard clause pattern replaces nested if statements
- Early returns for cleaner control flow
- Type hints for better documentation
- Proper error handling and cleanup
- More readable and testable code structure
Features:
- Starts Xvfb and web-ext
- Waits for extension installation
- Checks for JavaScript errors
- Verifies debugger connectivity
- Clean process termination
Usage: scripts/test_verify.py
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Translated Polish code comments in:
- components/report-window/problems/unlawful-cookies.tsx
- lib/browser-api/index.ts
- lib/browser-api/firefox.ts
- lib/browser-api/chrome.ts
- lib/browser-api/types.ts
Note: UI strings remain in Polish as per project language policy
(extension is designed for Polish users)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added Makefile section with verify, clean, and help targets
- Updated docker-compose commands to use 'docker compose' (v2 syntax)
- Added rentgen_verify service documentation
- Marked Makefile as recommended for testing
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Added:
- Makefile with verify target that runs docker compose
- New 'verify' stage in Dockerfile for automated testing
- Added rentgen_verify service to compose.yml
- Verification exits with proper exit code (0=success, non-zero=failure)
The make verify command:
1. Builds extension with docker compose
2. Runs test_verify.sh in headless Firefox
3. Propagates exit code from verification script
4. Fails if no proof of code execution found
Usage: make verify
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- verify_extension_working.sh: Polish → English
- verify_extension_code.sh: Polish → English
- functional_test.sh: Polish → English
- All 5 bash scripts now fully in English
- No functional changes, comments and echo messages only
- Dockerfile header reduced to single line reference to README
- Added comprehensive 'Docker Usage' section to README.md
- Easier to maintain documentation in one place
- Includes all usage examples: build, test, runtime, docker-compose
- usunięto wszystkie console.error z memory.ts (user widział czerwony tekst)
- weryfikacja teraz przez:
1. Sprawdzenie braku błędów JS
2. Sprawdzenie dostępności Remote Debugging Protocol
3. Dowód: extension zainstalowany + background page załadowana + brak błędów
- to jest NIEINWAZYJNE - nie psuje UX produkcji
- badge API to część normalnej operacji (nie test-only code)
- console.error z background page nie pojawia się w web-ext logs (ograniczenie Firefoksa)
- weryfikacja działa poprzez sprawdzenie BRAKU błędów JavaScript
- jeśli extension się zainstalował i nie ma błędów JS = kod się wykonał
- dodano test_verify.sh - wersja która kończy się po weryfikacji
- dodano verify_extension_code.sh i functional_test.sh dla future use
Dodano console.error logi w kluczowych punktach extensiona:
- Inicjalizacja extensiona (init())
- Konstruktor Memory (setup webRequest listeners)
- Pierwszy przechwycony request
**Uwaga:** Te logi są widoczne tylko w Firefox Browser Console
(Ctrl+Shift+J), nie w web-ext stdout. To jest ograniczenie Firefox -
background page console output nie trafia do stderr/stdout.
Zaktualizowano też skrypt test_start_extension.sh aby szukał logów
[RENTGEN] w przypadku gdy byłyby widoczne.
Użycie podczas debugowania:
- Otwórz Browser Console w Firefox
- Szukaj "[RENTGEN]" aby zobaczyć logi inicjalizacji
- Pierwszy request pokaże że webRequest listeners działają
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Utworzono scripts/test_start_extension.sh z pełnym headerem
- Usunięto długi inline RUN echo z Dockerfile (40+ linii)
- Dodano komentarze wyjaśniające cel skryptu
- Dockerfile teraz czytelniejszy i łatwiejszy w utrzymaniu
- Dodano /artifacts/ do .gitignore (docker buildx output)
Skrypt zawiera:
- Wyjaśnienie czemu istnieje (header)
- Uruchomienie Xvfb i web-ext
- Weryfikację instalacji extensiona
- Czytelne komunikaty sukcesu
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Skrypt sprawdza czy extension rzeczywiście się zainstalował
- Wyświetla czytelny komunikat sukcesu z info o procesach
- Capture output web-ext do /tmp/web-ext.log dla weryfikacji
- Timeout 30s na instalację extensiona
- Pokazuje PID Firefox i Xvfb
Weryfikacja pokazuje:
✓ Extension installed: rentgen@internet-czas-dzialac.pl
✓ Firefox ESR running in headless mode
✓ Xvfb running on display :99
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Dodano .claude do .gitignore (katalog Claude Code)
- Aktualizacja package-lock.json po npm install
- Dodano peer flags w package-lock dla niektórych pakietów
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Dodano nowy stage 'runtime' w Dockerfile
- Instalacja Firefox ESR i Xvfb dla headless execution
- Automatyczne uruchomienie web-ext run z Xvfb
- Dodano usługę rentgen_run w compose.yml
- Zaktualizowana dokumentacja z przykładami użycia
Możliwe użycie:
- docker build --target runtime -t rentgen-run .
- docker run --rm -it rentgen-run
- docker compose up rentgen_run
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Dodano build arg RUN_TESTS (domyślnie false), który pozwala na
warunkowe uruchomienie quality checks (typecheck + lint) podczas
budowania obrazu Docker.
Użycie:
- Bez testów: docker build -t rentgen .
- Z testami: docker build --build-arg RUN_TESTS=true -t rentgen .
Testy dodają ~10 sekund do czasu buildu.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Zmiana typu zwracanego w use-survey.ts z Survey.ReactSurveyModel na Survey.Model
- Dodanie brakującej prop refreshToken w StolenDataCluster
Typecheck teraz przechodzi bez błędów.
- Dodano Dockerfile z multi-stage build (artifacts + dev environment)
- Dodano .dockerignore dla optymalizacji budowania
- Dodano CLAUDE.md z dokumentacją architektury i workflow dla Claude Code
