Windows: test_platform.test_architecture_via_symlink() regression

[vstinner]

BPO	28059
Nosy	@pfmoore, @vstinner, @tjguk, @zware, @zooba

^{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/zooba'
closed_at = <Date 2016-09-10.19:25:26.597>
created_at = <Date 2016-09-10.08:13:18.315>
labels = ['OS-windows']
title = 'Windows: test_platform.test_architecture_via_symlink() regression'
updated_at = <Date 2016-09-10.19:25:26.596>
user = 'https://github.com/vstinner'

bugs.python.org fields:

activity = <Date 2016-09-10.19:25:26.596>
actor = 'python-dev'
assignee = 'steve.dower'
closed = True
closed_date = <Date 2016-09-10.19:25:26.597>
closer = 'python-dev'
components = ['Windows']
creation = <Date 2016-09-10.08:13:18.315>
creator = 'vstinner'
dependencies = []
files = []
hgrepos = []
issue_num = 28059
keywords = []
message_count = 7.0
messages = ['275596', '275651', '275659', '275661', '275663', '275682', '275683']
nosy_count = 6.0
nosy_names = ['paul.moore', 'vstinner', 'tim.golden', 'python-dev', 'zach.ware', 'steve.dower']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = None
url = 'https://bugs.python.org/issue28059'
versions = []

[vstinner]

It looks like test_platform started to fail on AMD64 Windows10 3.x between build 1456 (success) and 1458 (failure, 1457 was interrupted or something like that).

The regression may be related to the issue bpo-27932, change 31b7eaff5588.

http://buildbot.python.org/all/builders/AMD64%20Windows10%203.x/builds/1458/steps/test/logs/stdio

======================================================================
FAIL: test_architecture_via_symlink (test.test_platform.PlatformTest)
----------------------------------------------------------------------

Traceback (most recent call last):
  File "D:\buildarea\3.x.bolen-windows10\build\lib\test\test_platform.py", line 33, in test_architecture_via_symlink
    self.assertEqual(get(real), get(link))
AssertionError: Tuples differ: (b"('64bit', 'WindowsPE')\r\n", None) != (b'', None)

First differing element 0:
b"('64bit', 'WindowsPE')\r\n"
b''

• (b"('64bit', 'WindowsPE')\r\n", None)
  + (b'', None)

----------------------------------------------------------------------

[vstinner]

http://buildbot.python.org/all/builders/AMD64%20Windows8%203.x/builds/2553/steps/test/logs/stdio

test_architecture_via_symlink (test.test_platform.PlatformTest) ... Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "D:\buildarea\3.x.bolen-windows8\build\lib\platform.py", line 501, in <module>
    import ctypes
  File "D:\buildarea\3.x.bolen-windows8\build\lib\ctypes\__init__.py", line 7, in <module>
    from _ctypes import Union, Structure, Array
ModuleNotFoundError: No module named '_ctypes'
FAIL

[zooba]

That's strange, win32 should always have _ctypes available. I'll try and take a look later today

[zooba]

Though ctypes here is incorrect anyway. I really ought to rewrite it into the winapi module.

I'll revert the memory leak fix and fix the tests for b1, then look at converting it into a native function for b2.

[vstinner]

Though ctypes here is incorrect anyway. I really ought to rewrite it into the winapi module.

I agree, it's much better to avoid ctypes in the stdlib.

[zooba]

This bug is entirely because _ctypes cannot be found when we launch from the symlink without also setting PYTHONPATH.

I've already removed the import of _ctypes, but I'll also set PYTHONPATH in the test as there's no reason to require only builtins here.

[python-dev]

New changeset 94563ec74e1d by Steve Dower in branch '3.5':
Closes bpo-28059: Fixes test_platform to set PYTHONPATH for .pyd files
https://hg.python.org/cpython/rev/94563ec74e1d

New changeset 3ec4feb52a5b by Steve Dower in branch 'default':
Closes bpo-28059: Fixes test_platform to set PYTHONPATH for .pyd files
https://hg.python.org/cpython/rev/3ec4feb52a5b