feat(browser_manager): enable full XPI cleanup

Author: ndanner-wesleyancs

openwpm/browser_manager.py

@@ -158,6 +158,25 @@ def check_queue(launch_status: Dict[str, bool]) -> Any:
             # Resets the command/status queues
             (self.command_queue, self.status_queue) = (Queue(), Queue())
 
+            # Create a unique temporary directory for this spawn attempt.
+            # This is created inside the loop so each retry gets a fresh
+            # tmpdir (close_browser_manager deletes it on failure).
+            if self.browser_params.tmpdir is not None:
+                self.logger.debug(
+                    "BROWSER %i: Deleting leftover temp directory %s"
+                    % (self.browser_id, self.browser_params.tmpdir)
+                )
+                shutil.rmtree(self.browser_params.tmpdir, ignore_errors=True)
+            self.browser_params.tmpdir = Path(
+                tempfile.mkdtemp(
+                    prefix="openwpm_", dir=os.getenv("TMPDIR", default="/tmp")
+                )
+            )
+            self.logger.debug(
+                "BROWSER %i: Using temp dir %s"
+                % (self.browser_id, self.browser_params.tmpdir)
+            )
+
             # builds and launches the browser_manager
 
             self.browser_manager = BrowserManager(
@@ -341,6 +360,22 @@ def close_browser_manager(self, force: bool = False) -> None:
             if not shutdown_complete:
                 self.kill_browser_manager()
 
+            # Delete the temporary directory used by geckodriver.
+            if self.browser_params.tmpdir is not None:
+                try:
+                    shutil.rmtree(self.browser_params.tmpdir)
+                    self.logger.debug(
+                        "BROWSER %i: Deleted temp dir %s"
+                        % (self.browser_id, self.browser_params.tmpdir)
+                    )
+                    self.browser_params.tmpdir = None
+                except Exception:
+                    self.logger.warning(
+                        "BROWSER %i: Failed to delete temp dir %s"
+                        % (self.browser_id, self.browser_params.tmpdir),
+                        exc_info=True,
+                    )
+
     def execute_command_sequence(
         self,
         # Quoting to break cyclic import, see https://stackoverflow.com/a/39757388

openwpm/config.py

@@ -103,6 +103,7 @@ class BrowserParams(DataClassJsonMixin):
         default=Path(tempfile.gettempdir()),
         metadata=DCJConfig(encoder=path_to_str, decoder=str_to_path),
     )
+
     """
     The tmp_profile_dir defaults to the OS's temporary file folder (typically /tmp) and is where the generated 
     browser profiles and residual files are stored.
@@ -139,6 +140,18 @@ class BrowserParams(DataClassJsonMixin):
 
     """
 
+    tmpdir: Optional[Path] = field(
+        default=None,
+        metadata=DCJConfig(encoder=path_to_str, decoder=str_to_path),
+    )
+    """
+    The temporary directory used by `geckodriver`.  This is configured in
+    `BrowserManager.run` and then deleted when the browser is finished.  We do
+    this because it seems that `geckodriver` doesn't clean up its temporary
+    files (in particular, a copy of the extension XPI file), so we need to do
+    so ourselves.
+    """
+
     recovery_tar: Optional[Path] = None
     donottrack: bool = False
     tracking_protection: bool = False

openwpm/deploy_browsers/deploy_firefox.py

@@ -140,6 +140,15 @@ def deploy_firefox(
     # Launch the webdriver
     status_queue.put(("STATUS", "Launch Attempted", None))
 
+    # Use browser_params.tmpdir as the temporary directory.  This is so that
+    # geckodriver makes its copy of the extension XPI file in tmpdir, so
+    # we can delete it later and not have it left behind.  I make a shallow
+    # copy of `os.environ` because I'm a little nervous about modifying the
+    # OpenWPM process' environment.
+    env = os.environ.copy()
+    if browser_params.tmpdir is not None:
+        env["TMPDIR"] = str(browser_params.tmpdir)
+
     fo.binary_location = firefox_binary_path
     geckodriver_path = subprocess.check_output(
         "which geckodriver", encoding="utf-8", shell=True
@@ -149,6 +158,7 @@ def deploy_firefox(
         service=Service(
             executable_path=geckodriver_path,
             log_output=open(webdriver_interceptor.fifo, "w"),
+            env=env,
         ),
     )