html2print: minor cleanups on windows

Author: thseiler

.gitattributes

@@ -0,0 +1,3 @@
+# Mark the file as binary, so that Git won't not convert the line endings.
+# Otherwise, the file becomes unreadable for bash inside Docker (on Windows platforms).
+entrypoint.sh -text

.github/workflows/ci-windows.yml

@@ -28,18 +28,7 @@ jobs:
       shell: powershell
 
     - name: Check Chrome Version
-      run: '& "C:\Program Files\Google\Chrome\Application\chrome.exe" --version'
-      shell: powershell
-
-    - name: Add Chrome to PATH
-      run: |
-        $chromePath = "C:\Program Files\Google\Chrome\Application"
-        echo "Adding $chromePath to PATH"
-        echo "$chromePath" | Out-File -Append -Encoding utf8 $env:GITHUB_PATH
-      shell: powershell
-
-    - name: Verify Chrome Installation
-      run: chrome --version
+      run: '(Get-Item "C:\Program Files\Google\Chrome\Application\chrome.exe").VersionInfo'
       shell: powershell
 
     - name: Upgrade pip

html2print/html2print.py

@@ -39,7 +39,12 @@
 
 class ChromeDriverManager:
     def get_chrome_driver(self, path_to_cache_dir: str) -> str:
-        chrome_version: Optional[str] = self.get_chrome_version()
+        if platform.system() == "Windows":
+            import browsers
+
+            chrome_version: Optional[str] = browsers.get("chrome")["version"]
+        else:
+            chrome_version: Optional[str] = self.get_chrome_version()
 
         # If Web Driver Manager cannot detect Chrome, it returns None.
         if chrome_version is None:
@@ -109,7 +114,18 @@ def _download_chromedriver(
         path_to_cached_chrome_driver: str,
     ) -> str:
         url = "https://googlechromelabs.github.io/chrome-for-testing/known-good-versions-with-downloads.json"
-        response = ChromeDriverManager.send_http_get_request(url).json()
+        response = ChromeDriverManager.send_http_get_request(url)
+        if response is None:
+            raise RuntimeError(
+                "Could not download known-good-versions-with-downloads.json"
+            )
+
+        response = response.json()
+        if response is None:
+            raise RuntimeError(
+                "Could not parse known-good-versions-with-downloads.json"
+            )
+        assert isinstance(response, dict)
 
         matching_versions = [
             item
@@ -118,7 +134,7 @@ def _download_chromedriver(
         ]
 
         if not matching_versions:
-            raise Exception(
+            raise RuntimeError(
                 f"No compatible ChromeDriver found for Chrome version {chrome_major_version}"
             )
 
@@ -142,6 +158,11 @@ def _download_chromedriver(
         )
         response = ChromeDriverManager.send_http_get_request(driver_url)
 
+        if response is None:
+            raise Exception(
+                f"Could not download ChromeDriver from {driver_url}"
+            )
+
         Path(path_to_driver_cache_dir).mkdir(parents=True, exist_ok=True)
         zip_path = os.path.join(path_to_driver_cache_dir, "chromedriver.zip")
         print(  # noqa: T201
@@ -179,6 +200,9 @@ def send_http_get_request(url: str) -> Response:
             f"html2print: "
             f"failed to get response for URL: {url} with error: {last_error}"
         )
+        raise RuntimeError(
+            f"GET request failed after 3 attempts: {url}"
+        ) from last_error
 
     @staticmethod
     def get_chrome_version() -> Optional[str]:
@@ -328,14 +352,24 @@ def create_webdriver(
         service = Service(path_to_chrome)
 
     webdriver_options = Options()
+    # Workaround for Windows: chrome is not typically found in the PATH, so need to supply the exact binary location manually
+    if platform.system() == "Windows":
+        import browsers
+
+        chrome_binary_location = browsers.get("chrome")["path"]
+        webdriver_options.binary_location = chrome_binary_location
     webdriver_options.add_argument("start-maximized")
     webdriver_options.add_argument("disable-infobars")
     # Doesn't seem to be needed.
     # webdriver_options.add_argument('--disable-gpu')  # noqa: ERA001
     webdriver_options.add_argument("--disable-extensions")
-    webdriver_options.add_argument("--headless=chrome")
-    # FIXME: This is not nice but otherwise it does not work in Ubuntu 24-based Docker image.
-    # https://github.com/SeleniumHQ/selenium/issues/15327#issuecomment-2689287561
+    # Use --headless=new, as it seems to be more stable on Windows (available since Chrome 109).
+    # see https://www.selenium.dev/blog/2023/headless-is-going-away/
+    webdriver_options.add_argument("--headless=new")
+    # Docker disables some syscalls that are required for Chrome's sandbox to work.
+    # https://stackoverflow.com/questions/68855734/how-to-setup-chrome-sandbox-on-docker-container.
+    # See also https://github.com/SeleniumHQ/selenium/issues/15327#issuecomment-2689287561.
+    # We prefer isolation of the container over isolation of tabs within Chrome, and thus disable the sandbox.
     webdriver_options.add_argument("--no-sandbox")
 
     # The Chrome option --disable-dev-shm-usage disables the use of /dev/shm

pyproject.toml

@@ -54,6 +54,9 @@ dependencies = [
 
     # requests is used by HTML2PDF_HTTPClient.
     "requests",
+
+    # To detect the browers path (needed on Windows)
+    'pybrowsers; platform_system == "Windows"',
 ]
 
 [project.optional-dependencies]

tests/integration/lit.cfg

@@ -23,3 +23,6 @@ config.suffixes = ['.itest', '.c']
 config.is_windows = lit_config.isWindows
 if not lit_config.isWindows:
     config.available_features.add('PLATFORM_IS_NOT_WINDOWS')
+
+# $HOME is required for Chrome under Linux as it wants to store things in ~/.local
+config.environment['HOME'] = os.environ.get('HOME', '/tmp')