Found bug, fixed it, here is my solution....

[deeredman1991]

I already fixed the issue, so I don't need help, but I popped in to let you guys know there is an issue and how I solved it.

So I configured my vimrc to use python3

let g:powerline_pycmd = 'py3'

after that I started getting an exception in my vim console, so I output it to a logfile

This was the file:

2018-01-23 01:40:12,716:DEBUG:vim:Using libuv-based watcher
2018-01-23 01:40:16,876:ERROR:vim:branch:Exception while computing segment: _getvolumepathname() argument 1 must be str, not bytes
Traceback (most recent call last):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segment.py", line 173, in process_segment
    contents = segment['contents_func'](pl, segment_info)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segment.py", line 412, in <lambda>
    contents_func = lambda pl, segment_info: _contents_func(pl=pl, segment_info=segment_info, **args)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segments\common\vcs.py", line 21, in __call__
    repo = guess(path=name, create_watcher=create_watcher)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\lib\vcs\__init__.py", line 236, in guess
    for directory in generate_directories(path):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\lib\vcs\__init__.py", line 23, in generate_directories
    if os.path.ismount(path):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\ntpath.py", line 294, in ismount
    return path.rstrip(seps) == _getvolumepathname(path).rstrip(seps)
TypeError: _getvolumepathname() argument 1 must be str, not bytes
2018-01-23 01:40:16,877:ERROR:vim:file_vcs_status:Exception while computing segment: _getvolumepathname() argument 1 must be str, not bytes
Traceback (most recent call last):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segment.py", line 173, in process_segment
    contents = segment['contents_func'](pl, segment_info)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segment.py", line 412, in <lambda>
    contents_func = lambda pl, segment_info: _contents_func(pl=pl, segment_info=segment_info, **args)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segments\vim\__init__.py", line 555, in file_vcs_status
    repo = guess(path=name, create_watcher=create_watcher)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\lib\vcs\__init__.py", line 236, in guess
    for directory in generate_directories(path):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\lib\vcs\__init__.py", line 23, in generate_directories
    if os.path.ismount(path):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\ntpath.py", line 294, in ismount
    return path.rstrip(seps) == _getvolumepathname(path).rstrip(seps)
TypeError: _getvolumepathname() argument 1 must be str, not bytes

turns out it was an issue with python3 using bytes instead of strings for some things now...
So tl;dr I changed these lines of code and it fixed the issue on my end (blocks/lines I added are commented)

IN: .\powerline\lib\vcs\__init__.py:

def generate_directories(path):
	#Converts path from 'bytes' to 'str' object
	path = path.decode('utf-8')
	if os.path.isdir(path):
		#Converts path back into bytes and sends it out to wherever it needs to go.
		yield bytes(path, 'utf-8')
	while True:
		if os.path.ismount(path):
			break
		old_path = path
		path = os.path.dirname(path)
		if path == old_path or not path:
			break
		#Converts path back into bytes and sends it out to wherever it needs to go.
		yield bytes(path, 'utf-8')

IN: .\powerline\lib\shell.py:

def readlines(cmd, cwd):
	'''Run command and read its output, line by line

	:param list cmd:
		Command which will be run.
	:param str cwd:
		Working directory of the command which will be run.
	'''
	
	#Converts every 'bytes' object in cmd into a str object
	cmd = list(cmd)
	for i in range(len(cmd)):
		if type(cmd[i]) == type(bytes('', 'utf-8')):
			cmd[i] = cmd[i].decode()
	cmd = tuple(cmd)
	
	#converts cwd to 'str' object
	cwd = cwd.decode()
	
	p = Popen(cmd, shell=False, stdout=PIPE, stderr=PIPE, cwd=cwd)
	encoding = get_preferred_input_encoding()
	p.stderr.close()
	with p.stdout:
		for line in p.stdout:
			yield line[:-1].decode(encoding)

I'm sure there are more elegant ways of doing this (and that is why I am not issuing a pull request) but feel free to make any modifications to and/or use this solution in your codebase if you want. 👍

[ZyX-I]

It is not really valid to assume specific encoding of the paths: in linux file name is absolutely any sequence of bytes which does not include / (for it being a path separator) and NUL (for it terminating a C string). Command arguments are only not allowed to have NUL, but otherwise lack any encoding requirements. According to Python documentation, os.path.ismount must accept any path-like object, and “path-like” does include bytes. Same goes for Popen’s cwd. And though I don’t find any details of what kind of strings it accepts as an arguments, I do not know any code which will call that with non-unicode (non-str in Python-3*) strings.

[deeredman1991]

@ZyX-I While it is true that "os.path.ismount must accept any path-like object, and "path-like" does include bytes". If you notice; I modified the yield statements in generate_directories to cast path from a string to a byte. The original code, when I first opened the file, was:

def generate_directories(path):
	if os.path.isdir(path):
		yield path
	while True:
		if os.path.ismount(path):
			break
		old_path = path
		path = os.path.dirname(path)
		if path == old_path or not path:
			break
		yield path

^Copied from Github repository.

The issue was that path was a string and os.path.ismount was expecting a byte(or other "path-like" object I assume).

Also, (and I could be wrong on this) but I think I tried casting cmd[0:-1] and cwd to bytes first and that didn't work so I casted it all to strings and that fixed the issue. :)

[ZyX-I]

“Path-like object” both applies to str() and bytes(), so no matter what generate_directories actually accepted of these two it should actually work, and the code should be fine as long as the caller knows how to deal with generate_directories() returning the same type of object it was called with.

[deeredman1991]

@ZyX-I Oh wow, yeah, I just tested it, you are absolutely right. os.path.ismount() takes bytes AND strings...My question is then; why did python throw an error? and more importantly...how did my code fix it? (⊙ヘ⊙ ; )

[deeredman1991]

Yeah...I just reverted the code back to the way it was and threw a print statement in ntpath.py

try:
    from nt import _getvolumepathname
except ImportError:
    _getvolumepathname = None
def ismount(path):
    """Test whether a path is a mount point (a drive root, the root of a
    share, or a mounted volume)"""
    path = os.fspath(path)
    seps = _get_bothseps(path)
    path = abspath(path)
    root, rest = splitdrive(path)
    
    #I added this line...
    print('{0} is of type {1}'.format( path, type(path) ))
    
    if root and root[0] in seps:
        return (not rest) or (rest in seps)
    if rest in seps:
        return True
        
    if _getvolumepathname:
        return path.rstrip(seps) == _getvolumepathname(path).rstrip(seps)
    else:
        return False

This is what I got out:

b'J:\\PCTools\\cmder_mini\\test.py' is of type <class 'bytes'>
2018-01-26 11:33:20,291:ERROR:vim:branch:Exception while computing segment: _getvolumepathname() argument 1 must be str, not bytes
Traceback (most recent call last):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segment.py", line 173, in process_segment
    contents = segment['contents_func'](pl, segment_info)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segment.py", line 412, in <lambda>
    contents_func = lambda pl, segment_info: _contents_func(pl=pl, segment_info=segment_info, **args)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segments\common\vcs.py", line 21, in __call__
    repo = guess(path=name, create_watcher=create_watcher)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\lib\vcs\__init__.py", line 236, in guess
    for directory in generate_directories(path):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\lib\vcs\__init__.py", line 24, in generate_directories
    if os.path.ismount(path):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\ntpath.py", line 297, in ismount
    return path.rstrip(seps) == _getvolumepathname(path).rstrip(seps)
TypeError: _getvolumepathname() argument 1 must be str, not bytes
b'J:\\PCTools\\cmder_mini\\test.py' is of type <class 'bytes'>
2018-01-26 11:33:20,294:ERROR:vim:file_vcs_status:Exception while computing segment: _getvolumepathname() argument 1 must be str, not bytes
Traceback (most recent call last):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segment.py", line 173, in process_segment
    contents = segment['contents_func'](pl, segment_info)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segment.py", line 412, in <lambda>
    contents_func = lambda pl, segment_info: _contents_func(pl=pl, segment_info=segment_info, **args)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\segments\vim\__init__.py", line 555, in file_vcs_status
    repo = guess(path=name, create_watcher=create_watcher)
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\lib\vcs\__init__.py", line 236, in guess
    for directory in generate_directories(path):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\site-packages\powerline\lib\vcs\__init__.py", line 24, in generate_directories
    if os.path.ismount(path):
  File "J:\ProgrammingStuff\ProgrammingTools\Python36\lib\ntpath.py", line 297, in ismount
    return path.rstrip(seps) == _getvolumepathname(path).rstrip(seps)
TypeError: _getvolumepathname() argument 1 must be str, not bytes
b'J:\\PCTools\\cmder_mini\\test.py' is of type <class 'bytes'>
b'J:\\PCTools\\cmder_mini\\test.py' is of type <class 'bytes'>
b'J:\\PCTools\\cmder_mini\\test.py' is of type <class 'bytes'>
b'J:\\PCTools\\cmder_mini\\test.py' is of type <class 'bytes'>

so as you can see it does confirm that b'J:\\PCTools\\cmder_mini\\test.py' is of type <class 'bytes'> before it throws the error. So I have no idea...it might be possible that we just stumbled upon a bug in python...that...or there is a problem somewhere on my end... (⊙_⊙ ; )

I tried to find the nt module and add some print statements to _getvolumepathname() but it just keeps telling me that nt is a built in module...I guess that probably means it's a binary somewhere in my Python36 directory...There isn't a whole lot of info on it on Google...if you know how I can take a look at it; let me know and I'll go try some more print debugging.

I guess for now I'll just uncomment out my code and forget about it...at least until my PC explodes...lol</nervous laughter>

[ZyX-I]

“built-in” here means that it is in C code, possible, but not necessary, in the Python library/executable itself. Based on what I know on writing C extensions it will just have some call like PyArg_ParseTuple and it is what throws. Note a note near the s format.

[ZyX-I]

And it appears that Python has placed its nt module into Modules/posixmodule.c, and I forgot about another variant of parsing: you can declare function as accepting a specific set of arguments. What throws should be some kind of generated code because function itself uses PyUnicode_AsUnicodeAndSize, which calls PyErr_BadArgument on non-unicode arguments and that may only throw error with less helpful message.

[ZyX-I]

https://bugs.python.org/issue32693