Message223366
| Author |
zach.ware |
| Recipients |
berker.peksag, brett.cannon, ezio.melotti, serhiy.storchaka, zach.ware |
| Date |
2014年07月17日.20:09:19 |
| SpamBayes Score |
-1.0 |
| Marked as misclassified |
Yes |
| Message-id |
<1405627760.3.0.397447180395.issue22002@psf.upfronthosting.co.za> |
| In-reply-to |
| Content |
Here's a patch against 3.4 implementing Serhiy's suggestion in msg223277 and taking it a step further, actually using test discovery in all of the test.test_* subpackages.
To reduce duplication, the patch adds a 'load_package_tests' function to test.support, which is then used in each of the subpackages' load_tests function.
test_json and test_tools should have no visible changes from this patch. test_asyncio and test_email do have slight differences, but only in verbosity level: pre-patch, `python -m test.test_(asyncio|email)` runs at verbosity=2 (support.run_unittest default); with patch, they run at verbosity=1 (unittest default). test_asyncio also reports one more skipped test on Windows, due to a module that raises SkipTest on import.
@Brett: test_importlib sees the most changes, and I'd like to be sure that things are as you expect them to be. It looks like all of the test_suite() stuff is unused leftovers from when "test_importlib" was "importlib.test" and that test_importlib has actually been relying on unittest discovery (but bypassing load_tests and thereby not working with `python -m unittest test.test_importlib`), but I'd like confirmation on that.
A nice bonus with this patch is that (for example) `python -m test.test_importlib.source` works, testing just the named test_importlib subpackage. |
|