Change os.access to check ACLs under Windows

[tjguk]

BPO	2528
Nosy	@pfmoore, @amauryfa, @tiran, @tjguk, @serhiy-storchaka, @eryksun
Files	issue2528.2.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 = 'https://github.com/tjguk'
closed_at = None
created_at = <Date 2008-04-01.14:19:18.303>
labels = ['3.8', '3.9', '3.10', 'type-feature', 'library', 'OS-windows']
title = 'Change os.access to check ACLs under Windows'
updated_at = <Date 2021-03-12.02:11:01.666>
user = 'https://github.com/tjguk'

bugs.python.org fields:

activity = <Date 2021-03-12.02:11:01.666>
actor = 'eryksun'
assignee = 'tim.golden'
closed = False
closed_date = None
closer = None
components = ['Library (Lib)', 'Windows']
creation = <Date 2008-04-01.14:19:18.303>
creator = 'tim.golden'
dependencies = []
files = ['31165']
hgrepos = []
issue_num = 2528
keywords = ['patch']
message_count = 12.0
messages = ['64811', '109858', '109871', '109907', '110810', '192729', '194489', '194490', '236144', '243892', '243897', '388534']
nosy_count = 7.0
nosy_names = ['paul.moore', 'amaury.forgeotdarc', 'christian.heimes', 'tim.golden', 'piotr.dobrogost', 'serhiy.storchaka', 'eryksun']
pr_nums = []
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue2528'
versions = ['Python 3.8', 'Python 3.9', 'Python 3.10']

[tjguk]

At present, os.access under Windows simply calls GetFileAttributes to
determine the readonly attribute (ignoring directories). The patch
attached combines this with the use of the AccessCheck API to compare
the user's permissions with those required for the path.

I'm assuming that ATTRIB and CACLS will be available for use in the unit
tests included.

I haven't altered the structure of the posix_access function at all
although I suspect that it could now be simplified now that we're not
supporting Win9x.

[BreamoreBoy]

A quick look tells me that the patch seems clean. However it involves changes to posixmodule.c and I don't (yet) want to get involved with doing builds, so could someone else please give this a try.

[amauryfa]

But what are the benefits of this change?

[tjguk]

Although I'm the implementer of the patch (the concept
was discussed way back on c.l.py after a naive poster's
original request) I'm probably +0 myself. It's an attempt
to replace os.access' next-to-useless behaviour on Windows
with something which at least uses current security mechanisms
to determine its answer.

However, it's just one function among the many Posixesque
functions which don't quite map to Windows. And one which
you might well be advised to avoid in any case since there's
an race condition inherent in checking for a file's accessibility
and then making use of that fact. Better to try and fall back.

[BreamoreBoy]

Tim, do you want more time to think about this, could we close as "won't fix" or what?

[tiran]

Do you want to provide an updated patch for 3.4?

[tjguk]

Here's an updated patch against trunk with tests & doc changes

[tjguk]

... and to answer Amaury's question in msg109871 it creates a reasonable consistency between the results of os.access and the user's actual ability to read / write a file. eg, you might have no permissions whatsoever on the file but as long as it wasn't read-only, os.access would return True for reading, writing and executing.

[BreamoreBoy]

The solution proposed here could help resolve bpo-22107.