sys.remote_exec() doesn't work with non-UTF-8 paths

[serhiy-storchaka]

Bug report

There are two symptoms:

1. It does not work with non-ASCII names in non-UTF-8 locale. All related test_sys tests are failed on in non-UTF-8 locale. For example:

   $ LC_ALL=uk_UA ./python -m test -vuall test_sys -m test_remote_exec
   ======================================================================
   ERROR: test_remote_exec (test.test_sys.TestRemoteExec.test_remote_exec)
   Test basic remote exec functionality
   ----------------------------------------------------------------------
   Traceback (most recent call last):
     File "/home/serhiy/py/cpython/Lib/test/test_sys.py", line 2078, in test_remote_exec
       returncode, stdout, stderr = self._run_remote_exec_test(script)
                                    ~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^
     File "/home/serhiy/py/cpython/Lib/test/test_sys.py", line 2052, in _run_remote_exec_test
       sys.remote_exec(proc.pid, script_path)
       ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^
   FileNotFoundError: Script file does not exist
   
   ----------------------------------------------------------------------
   

2. Even in UTF-8 locale it fails if the path contains non-UTF-8 bytes.

The issue is twofold. On one side, sys.remote_exec() encodes the path to UTF-8 (even if the bytes path was decoded using the filesystem encoding). It fails because os.access() can't find file using the wrong path. It can accidentally success if other file exists with such path. On other side, PyFile_OpenFile() which interprets the path as UTF-8 is used to open the file.

Linked PRs

• gh-133886: Fix sys.remote_exec() for non-UTF-8 paths #133887
• [3.14] gh-133886: Fix sys.remote_exec() for non-UTF-8 paths (GH-133887) #133963