Setting breakpoint fails on Python files containing non-ASCII characters

[DonJayamanne]

Original issue can be found here microsoft/vscode-python#2914

[karthiknadig]

This is a 2.7 unicode file path issue. Relevant stack where this breaks:

C:\Users\Administrator\Desktop\项目源代码\unetfe-rotation>cd c:\Users\Administrator\Desktop\项目源代码\unetfe-rotation && cmd /C "set "PYTHONIOENCODING=UTF-8" && set "PYTHONUNBUFFERED=1" && python c:\Users\Administrator\.vscode\extensions\ms-python.python-2018.9.0\pythonFiles\experimental\ptvsd_launcher.py 53714 c:\Users\Administrator\Desktop\项目源代码\unetfe-rotation\app.py "
Traceback (most recent call last):
  File "c:\Users\Administrator\.vscode\extensions\ms-python.python-2018.9.0\pythonFiles\experimental\ptvsd\ptvsd\wrapper.py", line 1292, in done
    fut.result()
  File "c:\Users\Administrator\.vscode\extensions\ms-python.python-2018.9.0\pythonFiles\experimental\ptvsd\ptvsd\futures.py", line 40, in result
    reraise(self._exc_info)
  File "c:\Users\Administrator\.vscode\extensions\ms-python.python-2018.9.0\pythonFiles\experimental\ptvsd\ptvsd\futures.py", line 149, in callback
    x = next(it)
  File "c:\Users\Administrator\.vscode\extensions\ms-python.python-2018.9.0\pythonFiles\experimental\ptvsd\ptvsd\wrapper.py", line 2123, in on_setBreakpoints
    msg = '{}\t{}\t{}'.format(bp_type, path, vsc_bpid)
UnicodeEncodeError: 'ascii' codec can't encode characters in position 31-35: ordinal not in range(128)

We have a test, test_path_with_unicode this is disabled in 2.7, we should enable it after this is fixed.

[int19h]

The reason why this fails is because sys.getdefaultencoding() is ascii on Windows (regardless of locale), and that's what's used by implicit str<->unicode conversion in 2.7. In this case, path is actually unicode, but we're trying to convert to str because the string literal on the left is doesn't have the u prefix.

So far as I can see, just using u'...' form for the literal should fix this. We should probably do this for all places that format messages to sent to pydevd, anywhere in wrapper.py - basically any string that goes to pydevd_notify should be created as unicode (we can add an assert to that method to find all such places). The implementation of pydevd_notify already does proper encoding to bytes before sending it over.