So many things have been fixed since I last made a list of failing modules that it seemed like time to make a new one.  First, failing modules that already have open issues with patches:

- test_codecmaps* (issue18258)
- test_concurrent_futures (issue16968)
- test_locale (issue17767)
- test_multiprocessing (issue17778)
- test_socket (issue14408 -- incomplete patch, waiting on resolution of issue16968 for how to handle thread/process reaping)



Other tests that fail, that may be the fault of my test machine rather than the tests themselves (or, ones that I'd like others to try before I call them broken):

- test_urllib2net (fails on my machine on normal runs due to network setup)
- test_shutil (fails on my machine on normal runs for unknown reasons relating to symlinks)



The below tests fail for reasons that I expect to probably be related to test run order, as all of them pass when narrowing the pattern to match only them (e.g., "test_dbm*.py" rather than "test_*.py"):

- test_dbm (had a permission error on a test file)
- test_venv (tried to delete a temp dir that was unexpectedly non-empty)
- test_wsgiref (failure in testEnviron)
- test_logging (failure in test_warnings)
- test_inspect (had some issue with locating its test files)



This leaves only 6 other tests that fail due to improper inheritance, setup done in test_main rather than setUp/setUpClass/setUpModule, or other reasons directly related to discovery:

- test_decimal
- test_largefile
- test_pickle (pickletester)
- test_runpy
- test_shelve
- test_xml_etree



Other tests that need some manner of attention related to discovery:

- test_json: it half-uses discovery, it could use it completely
- test_importlib: about the same as test_json
- leakers package: either rename to non-discoverable names or otherwise make leakers.test_gestalt (in particular) not report an error when discovered.



This may not be a comprehensive list, but it does cover all of the most obvious failures that remain.