Subtle bug in os.path.realpath on Cygwin

[edemaine]

BPO	768419
Nosy	@loewis, @sjoerdmullender

^{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 = None
closed_at = <Date 2005-03-03.09:52:16.000>
created_at = <Date 2003-07-09.12:38:08.000>
labels = ['invalid', 'library']
title = 'Subtle bug in os.path.realpath on Cygwin'
updated_at = <Date 2005-03-03.09:52:16.000>
user = 'https://bugs.python.org/edemaine'

bugs.python.org fields:

activity = <Date 2005-03-03.09:52:16.000>
actor = 'loewis'
assignee = 'none'
closed = True
closed_date = None
closer = None
components = ['Library (Lib)']
creation = <Date 2003-07-09.12:38:08.000>
creator = 'edemaine'
dependencies = []
files = []
hgrepos = []
issue_num = 768419
keywords = []
message_count = 4.0
messages = ['16982', '16983', '16984', '16985']
nosy_count = 4.0
nosy_names = ['loewis', 'sjoerd', 'anadelonbrin', 'edemaine']
pr_nums = []
priority = 'normal'
resolution = 'not a bug'
stage = None
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue768419'
versions = []

[edemaine]

Cygwin allows mounting directories within other mount
points. This can cause os.path.realpath to expand some
symlinks that it shouldn't. For example:

$ cd /
$ mkdir X
$ mkdir X/Y
$ ln -s X Z
$ mount C:/ /Z/Y
$ ls Z/Y
[...contents of C:\...]
$ ls X/Y
[empty directory]

$ python -c "import os; print os.path.realpath('Z/Y')"
/X/Y
[bad because /X/Y is empty yet the original Z/Y has files]

In Cygwin, the correct answer would be either 'C:\\' or
'/cygdrive/c/'.

Conceivably this problem can happen in UNIces other
than Cygwin. It would be rather annoying to fix,
because it would require looking at the mount table.
But I thought I would mention it anyway...

[sjoerdmullender]

Logged In: YES
user_id=43607

I'd think this is rather a Cygwin bug than a Python bug.

Before the mount, /X/Y and /Z/Y refer to the same directory,
and on Unix they have the same device/inode combination.
And that combination is the all-important factor when doing
a mount on that directory.
If you do this on Unix (I used an NFS mount instead of the
mount shown in the report), both /X/Y and /Z/Y contain the
mounted directory.

[anadelonbrin]

Logged In: YES
user_id=552329

I agree with Sjoerd - this is a Cygwin bug, not a Python one.

[loewis]

Logged In: YES
user_id=21627

Closing as suggested.