| msg314340 - (view) |
Author: Hartmut Goebel (htgoebel) |
Date: 2018-03-23 23:30 |
| As of Python 3.7.0b2 _frozen_importlib_external.PathFinder exists twice on sys.meta_path, and it is the same object: $ python -S Python 3.7.0b2 (default, Mar 22 2018, 20:09:00) [GCC 5.5.0] on linux Type "help", "copyright", "credits" or "license" for more information. >>> import sys >>> print(sys.meta_path) [<class '_frozen_importlib.BuiltinImporter'>, <class '_frozen_importlib.FrozenImporter'>, <class '_frozen_importlib_external.PathFinder'>, <class '_frozen_importlib_external.PathFinder'>] >>> print([id(p) for p in sys.meta_path]) [24427944, 24430216, 24517416, 24517416] >>> |
|
|
| msg314348 - (view) |
Author: Ned Deily (ned.deily) *  |
Date: 2018-03-24 01:01 |
| Thanks for the report. git bisect result: 1abcf6700b4da6207fe859de40c6c1bada6b4fec is the first bad commit commit 1abcf6700b4da6207fe859de40c6c1bada6b4fec Author: Eric Snow <ericsnowcurrently@gmail.com> Date: Tue May 23 21:46:51 2017 -0700 bpo-22257: Private C-API for core runtime initialization (PEP 432). (#1772) (patch by Nick Coghlan) |
|
|
| msg314556 - (view) |
Author: Pablo Galindo Salgado (pablogsal) * |
Date: 2018-03-27 22:21 |
| It seems that the problem is that `_Py_InitializeEx_Private` calls `_Py_InitializeCore` and `_Py_InitializeMainInterpreter`. The first one calls at the end `initimport` that in turns calls `importlib._install_external_importers` that sets the PathFinder. Then, `_Py_InitializeMainInterpreter` calls `initexternalimport` that in turns calls again `importlib._install_external_importers` and therefore we end with two PathFinders in sys.meta_path. I think the solution is removing the call to initexternalimport in `_Py_InitializeMainInterpreter` as it has been alreade initialized by `_Py_InitializeCore`. |
|
|
| msg314557 - (view) |
Author: Pablo Galindo Salgado (pablogsal) * |
Date: 2018-03-27 22:33 |
| The same problem happens in `new_interpreter` as far as I understand. |
|
|
| msg314628 - (view) |
Author: Alyssa Coghlan (ncoghlan) * |
Date: 2018-03-29 01:50 |
| Calling initexternalimports InitalizeMainInterpreter and from new_interpreter is right, since we only want these importers on the metapath after we know sys.path has been configured appropriately. It's the call from InitializeCore that's questionable, since we haven't finished setting up sys.path at that point. |
|
|
| msg315717 - (view) |
Author: Alyssa Coghlan (ncoghlan) * |
Date: 2018-04-25 02:22 |
| New changeset 0977091dca59709864b14cfc129388f1f0de7bf7 by Nick Coghlan (Pablo Galindo) in branch 'master': bpo-33128 Fix duplicated call to importlib._install_external_importers (GH-6273) https://github.com/python/cpython/commit/0977091dca59709864b14cfc129388f1f0de7bf7 |
|
|
| msg315718 - (view) |
Author: miss-islington (miss-islington) |
Date: 2018-04-25 02:48 |
| New changeset 52a5a17338dfa7fed259027e1ecceba9c8491189 by Miss Islington (bot) in branch '3.7': bpo-33128 Fix duplicated call to importlib._install_external_importers (GH-6273) https://github.com/python/cpython/commit/52a5a17338dfa7fed259027e1ecceba9c8491189 |
|
|
| msg315719 - (view) |
Author: Alyssa Coghlan (ncoghlan) * |
Date: 2018-04-25 02:58 |
| Hartmut, thanks for the issue report, and Pablo, thanks for the PR to resolve it! |
|
|