> All filenames should use the filesystem encoding in Python.

Here is a new patch [code_encoding.patch] implementing this idea:

 - Use filesystem encoding (and surrogateescape) to encode/decode paths in compile() and the parser, instead of utf-8 in strict mode
 - Ensure that co_filename attribute can be used as a filename (eg. to not raise UnicodeEncodeError on Linux)
 - compile() builtin supports bytes filenames
 - _Py_FindSourceFile() (traceback.c) encodes paths of sys.path into the filesystem encoding, as do find_module() (import.c)
 - PyRun_SimpleFileExFlags() sets __file__ attribute using the filesystem encoding

The patch restores the situation before #6543.