gh-144503: Pass sys.argv as separate command line arguments.

The maximum length of a single command line argument is more restricted
than the total size of all command line arguments together.

Author: sjoerdjob

Lib/multiprocessing/forkserver.py

@@ -172,8 +172,6 @@ def ensure_running(self):
                     main_kws['sys_path'] = data['sys_path']
                 if 'init_main_from_path' in data:
                     main_kws['main_path'] = data['init_main_from_path']
-                if 'sys_argv' in data:
-                    main_kws['sys_argv'] = data['sys_argv']
                 if self._preload_on_error != 'ignore':
                     main_kws['on_error'] = self._preload_on_error
 
@@ -197,6 +195,8 @@ def ensure_running(self):
                     exe = spawn.get_executable()
                     args = [exe] + util._args_from_interpreter_flags()
                     args += ['-c', cmd]
+                    if self._preload_modules:
+                        args += data["sys_argv"]
                     pid = util.spawnv_passfds(exe, args, fds_to_pass)
                 except:
                     os.close(alive_w)
@@ -282,7 +282,7 @@ def _handle_preload(preload, main_path=None, sys_path=None, sys_argv=None,
 
 
 def main(listener_fd, alive_r, preload, main_path=None, sys_path=None,
-         *, sys_argv=None, authkey_r=None, on_error='ignore'):
+         *, authkey_r=None, on_error='ignore'):
     """Run forkserver."""
     if authkey_r is not None:
         try:
@@ -293,6 +293,11 @@ def main(listener_fd, alive_r, preload, main_path=None, sys_path=None,
     else:
         authkey = b''
 
+    if preload:
+        sys_argv = sys.argv[1:]
+    else:
+        sys_argv = None
+
     _handle_preload(preload, main_path, sys_path, sys_argv, on_error)
 
     util._close_stdin()

Lib/test/_test_multiprocessing.py

@@ -20,6 +20,7 @@
 import collections.abc
 import socket
 import random
+import resource
 import logging
 import shutil
 import subprocess
@@ -7106,6 +7107,28 @@ def test_preload_main_sys_argv(self):
             '',
         ])
 
+    def test_preload_main_sys_argv_limits(self):
+        # gh-144503: Check that sys.argv is set before __main__ is pre-loaded
+        if multiprocessing.get_start_method() != "forkserver":
+            self.skipTest("forkserver specific test")
+
+        max_str_arglen = 32 * resource.getpagesize()
+        argv = ["a" * (max_str_arglen - 1), "b"]
+        name = os.path.join(os.path.dirname(__file__), 'mp_preload_sysargv.py')
+        _, out, err = test.support.script_helper.assert_python_ok(
+            name, *argv)
+        self.assertEqual(err, b'')
+
+        out = out.decode().split("\n")
+        expected_argv = str(argv)
+        self.assertEqual(out, [
+            f"module:{expected_argv}",
+            f"fun:{expected_argv}",
+            f"module:{expected_argv}",
+            f"fun:{expected_argv}",
+            '',
+        ])
+
 #
 # Mixins
 #