| msg258333 - (view) |
Author: Brett Cannon (brett.cannon) *  |
Date: 2016-01-15 21:44 |
| Since loader.load_module() is documented as deprecated, we should consider raising an ImportWarning when it is used. That way when Python 2.7 support ends we can either remove it or have one more release where the various ImportWarnings turn into DeprecationWarnings and we finally clean up the import semantics to only be spec-based. |
|
|
| msg347561 - (view) |
Author: Joannah Nanjekye (nanjekyejoannah) * |
Date: 2019-07-09 13:29 |
| @brett Since this was opened like years ago. Before I work on it, I was wondering If it is still relevant. |
|
|
| msg347580 - (view) |
Author: Brett Cannon (brett.cannon) * |
Date: 2019-07-09 21:43 |
| @joannah no clue if this ever happened. Could you check the code if DeprecationWarning is being raised yet? |
|
|
| msg349284 - (view) |
Author: Michael Anckaert (michaelanckaert) * |
Date: 2019-08-09 13:25 |
| I checked out the source (Lib/imp.py:219) and only see the docstring marking this method als Deprecated. No exceptions are being raised. I would like to work on this issue. |
|
|
| msg349286 - (view) |
Author: Michael Anckaert (michaelanckaert) * |
Date: 2019-08-09 13:29 |
| Clarification: the imp module shows a DeprecationWarning when imported. Was this what was meant? |
|
|
| msg349287 - (view) |
Author: Joannah Nanjekye (nanjekyejoannah) * |
Date: 2019-08-09 13:30 |
| Feel free to open a PR. |
|
|
| msg349295 - (view) |
Author: Brett Cannon (brett.cannon) * |
Date: 2019-08-09 16:50 |
| If you look at https://github.com/python/cpython/blob/master/Lib/importlib/_bootstrap.py and https://github.com/python/cpython/blob/master/Lib/importlib/_bootstrap_external.py you will notice a bunch of comments near all the calls to load_module(). Those are the places that should be somehow triggering a DeprecationWarning. I think there's a compatibility shim function which would need to raise an exception so that anyone who has implemented load_module() as the only way to handle loading get a warning (but I think that requires certain loaders to have appropriate support which hopefully they have already but someone needs to double-check). After that every implementation of load_module() should have a warning. |
|
|
| msg379389 - (view) |
Author: Brett Cannon (brett.cannon) * |
Date: 2020-10-23 00:38 |
| A PR is now up. I ended up deprecating the load_module() methods in importlib itself and then raise ImportWarning in the import system itself when falling back to load_module(). |
|
|
| msg382548 - (view) |
Author: Brett Cannon (brett.cannon) * |
Date: 2020-12-04 23:39 |
| New changeset 2de5097ba4c50eba90df55696a7b2e74c93834d4 by Brett Cannon in branch 'master': bpo-26131: Deprecate usage of load_module() (GH-23469) https://github.com/python/cpython/commit/2de5097ba4c50eba90df55696a7b2e74c93834d4 |
|
|
| msg382796 - (view) |
Author: Brandt Bucher (brandtbucher) * |
Date: 2020-12-09 18:29 |
| I'm seeing the following test failure locally on master (doesn't seem to be showing up in CI though). I'm not too familiar with this stuff, but it looks related to this change: ====================================================================== FAIL: test_all (test.test___all__.AllTest) (module='distutils.command') ---------------------------------------------------------------------- Traceback (most recent call last): File "/home/bucher/src/cpython/Lib/test/test___all__.py", line 55, in check_all self.assertEqual(keys, all_set, "in module {}".format(modname)) File "/home/bucher/src/cpython/Lib/contextlib.py", line 140, in __exit__ next(self.gen) File "/home/bucher/src/cpython/Lib/test/support/warnings_helper.py", line 177, in _filterwarnings raise AssertionError("unhandled warning %s" % reraise[0]) AssertionError: unhandled warning {message : ImportWarning('VendorImporter.exec_module() not found; falling back to load_module()'), category : 'ImportWarning', filename : '<frozen importlib._bootstrap>', lineno : 681, line : None} ---------------------------------------------------------------------- Should we be ignoring the new warning in this test as well? I'm still not sure why I seem to be the only one seeing it. |
|
|
| msg382799 - (view) |
Author: Brett Cannon (brett.cannon) * |
Date: 2020-12-09 18:40 |
| What is "VendorImporter" (see the message of the ImportWarning)? That's not in the stdlib, so it looks like your system is injecting something via some `.pth` file or environment variable that doesn't define exec_module(). |
|
|
| msg382801 - (view) |
Author: Brandt Bucher (brandtbucher) * |
Date: 2020-12-09 19:10 |
| Yep, looks like that was it. Thanks! |
|
|
| msg382874 - (view) |
Author: Sviatoslav Sydorenko (webknjaz) * |
Date: 2020-12-11 17:07 |
| > What is "VendorImporter" @brett.cannon `VendorImporter` is something comping from `pkg_resources`. Because of an entangled traceback and the fact that it dynamically injects a vendored copy of `six`, it's hard to spot. Here's an upstream issue I filed about this warning: https://github.com/pypa/setuptools/issues/2481. |
|
|
| msg382986 - (view) |
Author: STINNER Victor (vstinner) * |
Date: 2020-12-14 13:06 |
| I wrote https://github.com/benjaminp/six/pull/343 to fix the six module. I would appreciate a review. |
|
|