PyImport_ExecCodeModuleObject not setting module attributes

[trevor3]

BPO	21226
Nosy	@brettcannon, @ncoghlan, @benjaminp, @ericsnowcurrently
Files	fix-PyImport_ExecCodeModuleObject.diff fix-PyImport_ExecCodeModuleObject.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/ericsnowcurrently'
closed_at = <Date 2014-05-13.00:31:35.113>
created_at = <Date 2014-04-14.21:04:27.103>
labels = ['type-bug', 'library']
title = 'PyImport_ExecCodeModuleObject not setting module attributes'
updated_at = <Date 2014-05-29.18:44:15.513>
user = 'https://bugs.python.org/trevor3'

bugs.python.org fields:

activity = <Date 2014-05-29.18:44:15.513>
actor = 'python-dev'
assignee = 'eric.snow'
closed = True
closed_date = <Date 2014-05-13.00:31:35.113>
closer = 'eric.snow'
components = ['Library (Lib)']
creation = <Date 2014-04-14.21:04:27.103>
creator = 'trevor3'
dependencies = []
files = ['34971', '34980']
hgrepos = []
issue_num = 21226
keywords = ['patch']
message_count = 7.0
messages = ['216217', '216850', '216864', '216886', '218161', '218391', '219356']
nosy_count = 7.0
nosy_names = ['brett.cannon', 'ncoghlan', 'benjamin.peterson', 'Arfrever', 'python-dev', 'eric.snow', 'trevor3']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue21226'
versions = ['Python 3.3', 'Python 3.4', 'Python 3.5']

[trevor3]

In Python/import.c, PyImport_ExecCodeModuleObject creates a new module object but doesn't set all of the attributes required for modules, such as __spec__ or __loader__.

This breaks mod_wsgi from 3.3 and up, which depends on PyImport_ExecCodeModuleEx, which delegates to PyImport_ExecCodeModuleObject for its module creation logic. See https://code.google.com/p/modwsgi/source/browse/mod_wsgi/mod_wsgi.c#6289

[ericsnowcurrently]

Here's a (currently segfaulting) patch that demonstrates how I'd like to solve this. Feedback on the approach is welcome. :) When I get a chance I'll debug the segfault.

[benjaminp]

The last argument to _PyObject_CallMethodIdObjArgs needs to be NULL...

[ericsnowcurrently]

Thanks, Benjamin. Here's an updated patch that passes all tests. It should be good to go. And yet again our zipimport implementation was a pain point. :(

[brettcannon]

LGTM although you forgot to use a loader instance instead of the class.

[python-dev]

New changeset 7d20e30bd540 by Eric Snow in branch '3.4':
Issue bpo-21226: Set all attrs in PyImport_ExecCodeModuleObject.
http://hg.python.org/cpython/rev/7d20e30bd540

New changeset bc324a49d0fc by Eric Snow in branch 'default':
Merge from 3.4 (for bpo-21226).
http://hg.python.org/cpython/rev/bc324a49d0fc

[python-dev]

New changeset baa7b5555656 by Eric Snow in branch '3.4':
Issue bpo-21226: fix a ref leak.
http://hg.python.org/cpython/rev/baa7b5555656

New changeset 57130574d1e8 by Eric Snow in branch 'default':
Issue bpo-21226: Merge from 3.4.
http://hg.python.org/cpython/rev/57130574d1e8