| msg204920 - (view) |
Author: Ronald Oussoren (ronaldoussoren) *  |
Date: 2013-12-01 14:11 |
| To reproduce: * create a package with the following structure: pkg/ __init__.py _sub.py * __init__.py contains: from pkg._sub import * * the contents of _sub.py is not important * in a python shell do: >>> import pkg._sub as s >>> import imp >>> imp.reload(s) Traceback (most recent call last): File "", line 1, in File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/imp.py", line 297, in reload return importlib.reload(module) File "/Library/Frameworks/Python.framework/Versions/3.4/lib/python3.4/importlib/__init__.py", line 123, in reload raise ImportError(_bootstrap._ERR_MSG.format(name), name=name) ImportError: No module named 'pkg._sub' >>> In earlier python versions this reloaded the pkg._sub module. Importing _sub from __init__ doesn't seem to be important. |
|
|
| msg205510 - (view) |
Author: Eric Snow (eric.snow) * |
Date: 2013-12-08 03:27 |
| This is actually a problem with importlib.reload() (which imp.reload() simply wraps). The attached patch provides a test that reproduces the error. I'll work on a fix ASAP. Interestingly, the kind of failure depends on frozen vs. source importlib: ====================================================================== ERROR: test_reload_submodule (test.test_importlib.test_api.Frozen_ReloadTests) ---------------------------------------------------------------------- Traceback (most recent call last): File "Lib/test/test_importlib/test_api.py", line 430, in test_reload_submodule self.init.reload(ham) File "Lib/importlib/__init__.py", line 161, in reload methods.exec(module) File "<frozen importlib._bootstrap>", line 1134, in exec AttributeError: 'NoneType' object has no attribute 'name' ====================================================================== ERROR: test_reload_submodule (test.test_importlib.test_api.Source_ReloadTests) ---------------------------------------------------------------------- Traceback (most recent call last): File "Lib/test/test_importlib/test_api.py", line 430, in test_reload_submodule self.init.reload(ham) File "Lib/importlib/__init__.py", line 158, in reload raise ImportError(msg.format(parent_name), name=parent_name) ImportError: parent 'spam' not in sys.modules ---------------------------------------------------------------------- |
|
|
| msg205511 - (view) |
Author: Eric Snow (eric.snow) * |
Date: 2013-12-08 03:51 |
| Actually, they're both getting the same error: AttributeError: 'NoneType' object has no attribute 'name' I forgot to clear the submodule from sys.modules first. |
|
|
| msg205517 - (view) |
Author: Eric Snow (eric.snow) * |
Date: 2013-12-08 05:46 |
| The problem was that importlib.reload() was not passing the parent's __path__ to importlib._bootstrap._find_spec(). This was a consequence of us restoring the pre-3.3 reload semantics. Patch attached. (Note to self: add Misc/NEWS entry) |
|
|
| msg205519 - (view) |
Author: Alyssa Coghlan (ncoghlan) * |
Date: 2013-12-08 06:21 |
| Patch looks reasonable to me. |
|
|
| msg205705 - (view) |
Author: Olivier Grisel (Olivier.Grisel) * |
Date: 2013-12-09 15:34 |
| I tested the patch on the current HEAD and it fixes a regression introduced between 3.3 and 3.4b1 that prevented to build scipy from source with "pip install scipy". |
|
|
| msg205767 - (view) |
Author: Roundup Robot (python-dev)  |
Date: 2013-12-10 03:05 |
| New changeset 1d67eb1df5a9 by Eric Snow in branch 'default': Issue 19851: Fix a regression in reloading submodules. http://hg.python.org/cpython/rev/1d67eb1df5a9 |
|
|