tempfile.gettempdir() didn't return the path with correct case.

[lygstate]

BPO	14255
Nosy	@birkenfeld, @ncoghlan, @pitrou, @tjguk, @briancurtin, @shimizukawa, @lygstate
Files	tempfile_normpath.patch issue14255.3.diff

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/tjguk'
closed_at = <Date 2013-10-28.10:32:31.212>
created_at = <Date 2012-03-12.01:12:42.243>
labels = ['type-bug', 'library']
title = "tempfile.gettempdir() didn't return the path with correct case."
updated_at = <Date 2013-10-28.10:32:31.211>
user = 'https://github.com/lygstate'

bugs.python.org fields:

activity = <Date 2013-10-28.10:32:31.211>
actor = 'tim.golden'
assignee = 'tim.golden'
closed = True
closed_date = <Date 2013-10-28.10:32:31.212>
closer = 'tim.golden'
components = ['Library (Lib)']
creation = <Date 2012-03-12.01:12:42.243>
creator = 'lygstate'
dependencies = []
files = ['24861', '32362']
hgrepos = []
issue_num = 14255
keywords = ['patch']
message_count = 20.0
messages = ['155424', '155717', '155736', '155745', '155769', '155861', '155875', '155892', '155949', '173307', '173426', '180190', '201125', '201126', '201147', '201168', '201169', '201277', '201480', '201520']
nosy_count = 10.0
nosy_names = ['georg.brandl', 'ncoghlan', 'pitrou', 'tim.golden', 'brian.curtin', 'mcjeff', 'jteh', 'shimizukawa', 'python-dev', 'lygstate']
pr_nums = []
priority = 'low'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue14255'
versions = ['Python 2.7', 'Python 3.3', 'Python 3.4']

[lygstate]

>>> print tempfile.gettempdir()
c:\users\dreamkxd\appdata\local\temp
>>>
And the real path is
C:\Users\dreamkxd\AppData\Local\Temp.

[mcjeff]

The actual implementation calls os.path.normcase from tempfile._get_default_tempdir. In this scenario here, that resolves to the ntpath module & triggers a lowercase conversion.

On the other hand, the posixpath module is simply an identity function.

Now, it *is* possible to force a Windows file system to run in a case-sensitive configuration. Since ntpath.py forces lower case, this could actually cause breakage in that situation, however rare.

In my opinion, changing the ntpath.normcase function to retain case probably yields more correct behavior.

[birkenfeld]

There's not much sense in having normcase() if it never does anything :)

But in this case, I don't really see why tempfile needs to use normcase().

[pitrou]

+1 for fixing.

[mcjeff]

The normcase call isn't exactly a no-op in ntpath.py as it also swaps / for \\. Removing the .lower() seems to simply make it a watered down version of normpath.

There are a couple of options I guess. We could simply keep the altsep, or update _get_default_tempdir to use normpath?

I'd be happy to do it once everyone agrees =)

[lygstate]

+1
_get_default_tempdir to use normpath

[mcjeff]

Here's a tiny patch that just changes normcase->normpath. This fixes the casing issue at the 'gettempdir' level, though it doesn't address the 'normcase' function itself.

Note that *both* macpath.py and ntpath.py use <s>.lower, which obviously won't fly on case-sensitive mounts. If someone more experienced than I has a suggestion on how to handle that -- or whether its worthwhile to even touch -- I'll gladly implement.

[lygstate]

It's possible to getting actual file name under Windows.
http://stackoverflow.com/questions/74451/getting-actual-file-name-with-proper-casing-on-windows

[mcjeff]

It doesn't seem to me that the right approach on either platform is to simply downcase. Maybe just deprecate the call altogether as its not doing anything normpath isn't if the <s>.lower call is removed?

[jteh]

This issue is much nastier than it seems when dealing with character sets that contain multi-byte characters in Python 2.7 on Windows. For example, on a Japanese system:

1. The ANSI code page is cp932 and Python 2.7 will return the TEMP environment variable as a str with this encoding.
2. If a user has a username containing the character \u5c71, this will be encoded as \x8eR, so the TEMP environment variable will include this character.
3. When normcase is called, \x8eR is converted to \x8er.
4. Because \x8eR is a multi-byte character, \x8er is an unrelated character; case insensitivity for "R" doesn't apply.
5. The result is that the correct temp directory is skipped. In fact, on most systems, IOError is raised because none of the other candidate directories are valid either.

[mcjeff]

Yeah clearly the wrong behavior on Winders.

I think that moving to 'normpath' instead of 'normcase' is likely the right thing to do. Patch is attached, so if someone with commit powers could review real quick I'll address whatever needs to be addressed.

IMHO, the 'normcase' calls are of questionable benefit since the posix call is simply an identity and as this bug details, it isn't necessarily the right behavior on Win.