Coverage mislabeling `async for`

[Skelmis]

Describe the bug
When attempting to generate coverage on an async for statement it says the loop did not complete, when it has, as indicated by the rest of the coverage.

        async for guild in self.cache.get_all_guilds():
            # Loop code here

        await self.cache.drop()

This results in await self.cache.drop() being marked green / run, but also getting a partial error with the async for 'not completing'. This results in

> Line xxx didn't jump to line yyy, because the loop on line xxx didn't complete

To Reproduce
How can we reproduce the problem? Please be specific. Don't just link to a failing CI job. Answer the questions below:

1. What version of Python are you using?
2. 3.9.7
3. What version of coverage.py are you using? The output of coverage debug sys is helpful.
4. -- sys -------------------------------------------------------
   version: 5.5
   coverage: /home/skelmis/Python/DPY-Anti-Spam/venv/lib/python3.9/site-packages/coverage/init.py
   tracer: -none-
   CTracer: available
   plugins.file_tracers: -none-
   plugins.configurers: -none-
   plugins.context_switchers: -none-
   configs_attempted: .coveragerc
   configs_read: /home/skelmis/Python/DPY-Anti-Spam/.coveragerc
   config_file: /home/skelmis/Python/DPY-Anti-Spam/.coveragerc
   config_contents: b'[run]\nbranch = True\nomit =\n antispam/abc.py\n antispam/init.py\n antispam/caches/init.py\n antispam/caches/memory/init.py\n antispam/caches/redis/init.py\n antispam/dataclasses/init.py\n antispam/enums/init.py\n antispam/plugins/init.py'
   data_file: -none-
   python: 3.9.7 (default, Sep 3 2021, 06:18:44) [GCC 10.3.0]
   platform: Linux-5.10.0-kali9-amd64-x86_64-with-glibc2.31
   implementation: CPython
   executable: /home/skelmis/Python/DPY-Anti-Spam/venv/bin/python
   def_encoding: utf-8
   fs_encoding: utf-8
   pid: 11297
   cwd: /home/skelmis/Python/DPY-Anti-Spam
   path: /home/skelmis/Python/DPY-Anti-Spam/venv/bin
   /usr/lib/python39.zip
   /usr/lib/python3.9
   /usr/lib/python3.9/lib-dynload
   /home/skelmis/Python/DPY-Anti-Spam/venv/lib/python3.9/site-packages
   environment: -none-
   command_line: /home/skelmis/Python/DPY-Anti-Spam/venv/bin/coverage debug sys
   sqlite3_version: 2.6.0
   sqlite3_sqlite_version: 3.36.0
   sqlite3_temp_store: 0
   sqlite3_compile_options: COMPILER=gcc-10.2.1 20210110
   ENABLE_COLUMN_METADATA
   ENABLE_DBSTAT_VTAB
   ENABLE_FTS3
   ENABLE_FTS3_PARENTHESIS
   ENABLE_FTS3_TOKENIZER
   ENABLE_FTS4
   ENABLE_FTS5
   ENABLE_JSON1
   ENABLE_LOAD_EXTENSION
   ENABLE_MATH_FUNCTIONS
   ENABLE_PREUPDATE_HOOK
   ENABLE_RTREE
   ENABLE_SESSION
   ENABLE_STMTVTAB
   ENABLE_UNLOCK_NOTIFY
   ENABLE_UPDATE_DELETE_LIMIT
   HAVE_ISNAN
   LIKE_DOESNT_MATCH_BLOBS
   MAX_SCHEMA_RETRY=25
   MAX_VARIABLE_NUMBER=250000
   OMIT_LOOKASIDE
   SECURE_DELETE
   SOUNDEX
   SYSTEM_MALLOC
   TEMP_STORE=1
   THREADSAFE=1
   USE_URI
5. What versions of what packages do you have installed? The output of pip freeze is helpful.
6. aiohttp==3.7.4.post0
   alabaster==0.7.12
   appdirs==1.4.4
   async-timeout==3.0.1
   attrs==21.2.0
   Babel==2.9.1
   backports.entry-points-selectable==1.1.0
   black==21.5b1
   bleach==4.1.0
   certifi==2021.5.30
   cffi==1.14.6
   cfgv==3.3.1
   chardet==4.0.0
   charset-normalizer==2.0.6
   click==8.0.1
   colorama==0.4.4
   colorlog==6.4.1
   coverage==5.5
   coveralls==3.1.0
   cryptography==3.4.8
   discord==1.0.1
   discord.py==1.7.3
   distlib==0.3.2
   docopt==0.6.2
   docutils==0.16
   filelock==3.0.12
   fuzzywuzzy==0.18.0
   hikari==2.0.0.dev101
   hypothesis==6.21.6
   identify==2.2.14
   idna==3.2
   imagesize==1.2.0
   importlib-metadata==4.8.1
   iniconfig==1.1.1
   jeepney==0.7.1
   Jinja2==3.0.1
   keyring==23.2.1
   MarkupSafe==2.0.1
   MouseInfo==0.1.3
   multidict==5.1.0
   mypy-extensions==0.4.3
   nest-asyncio==1.5.1
   nodeenv==1.6.0
   packaging==21.0
   pathspec==0.9.0
   pip-chill==1.0.1
   pkginfo==1.7.1
   platformdirs==2.3.0
   pluggy==1.0.0
   pre-commit==2.13.0
   pure25519==0.0.1
   py==1.10.0
   pycparser==2.20
   PyGetWindow==0.0.9
   Pygments==2.10.0
   PyMsgBox==1.0.9
   pyparsing==2.4.7
   pyperclip==1.8.2
   PyRect==0.1.4
   PyScreeze==0.1.27
   pytest==6.2.5
   pytest-asyncio==0.15.1
   python-Levenshtein==0.12.2
   python3-xlib==0.15
   PyTweening==1.0.3
   pytz==2021.1
   PyYAML==5.4.1
   readme-renderer==29.0
   regex==2021.8.28
   requests==2.26.0
   requests-toolbelt==0.9.1
   rfc3986==1.5.0
   SecretStorage==3.3.1
   six==1.16.0
   snakeviz==2.1.0
   snowballstemmer==2.1.0
   sortedcontainers==2.4.0
   Sphinx==4.2.0
   sphinx-rtd-theme==0.5.2
   sphinxcontrib-applehelp==1.0.2
   sphinxcontrib-devhelp==1.0.2
   sphinxcontrib-htmlhelp==2.0.0
   sphinxcontrib-jsmath==1.0.1
   sphinxcontrib-qthelp==1.0.3
   sphinxcontrib-serializinghtml==1.1.5
   toml==0.10.2
   tornado==6.1
   tqdm==4.62.2
   twine==3.4.1
   typed-ast==1.4.3
   typing-extensions==3.10.0.2
   urllib3==1.26.6
   virtualenv==20.8.0
   webencodings==0.5.1
   yarl==1.6.3
   zipp==3.5.0
7. What code are you running? Give us a specific commit of a specific repo that we can check out.
8. https://github.com/Skelmis/DPY-Anti-Spam/tree/rewrite
9. https://github.com/Skelmis/DPY-Anti-Spam/commit/44af0db7e48b8ae870fee32cbc80fc14c9c1ebc6
10. What commands did you run?
11. coverage run --source=antispam -m pytest
12. coverage html

Expected behavior
The line is marked as complete.

[nedbat]

I think this is a dup of #1158, but that also seemed to be fixed in Python 3.9.7. Are you sure that's the version of Python that shows the problem?

[nedbat]

I tested it: the problem does happen with Python 3.9.7, but does not happen with 3.10.0. On 3.10, line 894 in anti_spam_handler.py is marked as completely covered.

[Skelmis]

So is this going to be a won't fix, just upgrade kind of use case? As I am attempting to support multiple python versions.

[nedbat]

I don't think coverage.py can make this better, but I haven't tried to create a reproducible example for the CPython bug tracker yet, so I'm not sure.

You can mark the line as partial with # pragma: no branch if you want to silence the false positive.

[vytas7]

A variation of this happens for me even on 3.10.2: when breaking or returning from an async for loop, the alternative branch of not breaking but jumping back to the loop seems to be missed. It might be caused by genuinely missing tests, but I much doubt that because everything is covered under 3.8.x.

I can try to distill an MRE.

[vytas7]

I went on to file a new issue since it seems it only happens under the above mentioned conditions (not for any kind of async for completion as described in this issue for 3.9): #1324.

[nedbat]

Since this is fixed in CPython, I will close this issue.

[alicederyn]

I have seen a recurrence of this bug in CPython 3.9.18. In case any one else hits it, I figured out a workaround:

async for ...
    ...
else:
    pass

I'm not 100% clear on why this works — I tried adding it because I forgot how else worked and thought it was only run if the for loop is missed.