pathlib: Path.exists broken

[hvenev]

BPO	22759
Nosy	@birkenfeld, @pitrou, @hvenev
Files	py.patch path_exists_enotdir.patch path_exists_enotdir2.patch

^{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 2014-10-30.22:16:28.601>
created_at = <Date 2014-10-29.15:30:32.611>
labels = ['type-bug']
title = 'pathlib: Path.exists broken'
updated_at = <Date 2014-10-30.22:16:28.599>
user = 'https://github.com/hvenev'

bugs.python.org fields:

activity = <Date 2014-10-30.22:16:28.599>
actor = 'pitrou'
assignee = 'none'
closed = True
closed_date = <Date 2014-10-30.22:16:28.601>
closer = 'pitrou'
components = []
creation = <Date 2014-10-29.15:30:32.611>
creator = 'h.venev'
dependencies = []
files = ['37066', '37078', '37079']
hgrepos = []
issue_num = 22759
keywords = ['patch']
message_count = 11.0
messages = ['230214', '230221', '230222', '230225', '230292', '230293', '230294', '230295', '230297', '230303', '230304']
nosy_count = 4.0
nosy_names = ['georg.brandl', 'pitrou', 'python-dev', 'h.venev']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue22759'
versions = ['Python 3.4', 'Python 3.5']

[hvenev]

$ touch a
c: stat("a/x", ...) -> errno=ENOTDIR
$ python
>>> pathlib.Path('a/x').exists()

This should return False and not throw an exception.

Patch not tested.

[hvenev]

Tested and works.

[birkenfeld]

os.path.exists() ignores *all* OSErrors from stat. This is probably too broad, but only ignoring ENOENT is probably too narrow.

[hvenev]

ENAMETOOLONG and possibly ELOOP should also return False.
EACCES, EOVERFLOW and ENOMEM should probably be forwarded.
EFAULT should not happen.

[pitrou]

ENAMETOOLONG: that doesn't mean the file doesn't exist. For example, it could be accessed by changing the current directory and shortening the residual path name.

ELOOP: unfortunately it doesn't guarantee that there is a real symlink loop, just that the OS decided to bail out after a certain number of indirections.

So in both cases I think we should let the errors bubble up.

[pitrou]

Here is a patch with tests.

[hvenev]

Should I file bugs for is_dir, is_file, is_symlink, is_socket, is_fifo, is_block_device and is_char_device?

[pitrou]

Le 30/10/2014 20:36, Hristo Venev a écrit :

Hristo Venev added the comment:

Should I file bugs for is_dir, is_file, is_symlink, is_socket, is_fifo, is_block_device and is_char_device?

Hmm... good catch. No, this issue will do.

[pitrou]

Updated patch for other querying function.

[python-dev]

New changeset 40497cc445f4 by Antoine Pitrou in branch '3.4':
Issue bpo-22759: Query methods on pathlib.Path() (exists(), is_dir(), etc.) now return False when the underlying stat call raises NotADirectoryError.
https://hg.python.org/cpython/rev/40497cc445f4

New changeset ff5f5fd230d3 by Antoine Pitrou in branch 'default':
Issue bpo-22759: Query methods on pathlib.Path() (exists(), is_dir(), etc.) now return False when the underlying stat call raises NotADirectoryError.
https://hg.python.org/cpython/rev/ff5f5fd230d3

[pitrou]

This is now fixed. Thanks for the report!