AsyncioDispatcher cleanup tasks atexit

[mdavidsaver]

Currently, the atexit logic simply breaks the event loop leaving an in-progress Task incomplete. Instead, use asyncio.run(), which cancels tasks before returning (among other cleanup).

Motivated by usage of asyncio.create_subprocess_exec() where the present behavior leaves the child process running. Similarly, for cleanup of temporary files/directories.

The downside is that asyncio.run() is introduced in python 3.7. So the 3.6 tests will fail. Emulating asyncio.run() in 3.6 would be possible, but is it worth the effort?

I don't know what is going wrong with the windows and mac tests.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 88.15%. Comparing base (d55483d) to head (f7a1f53).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #138      +/-   ##
==========================================
+ Coverage   87.73%   88.15%   +0.41%     
==========================================
  Files          14       14              
  Lines        1019     1038      +19     
==========================================
+ Hits          894      915      +21     
+ Misses        125      123       -2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[coretl]

I think it is worth ditching 3.6 so we get asyncio.run. @AlexanderWells-diamond @Araneidae what do you think? This looks ok to me, I asked a couple of questions out of interest. It need 3.6 ditching from the matrix before merging. The mac and windows tests seem to be working in the latest run?

[AlexanderWells-diamond]

I have no attachment to Python 3.6 so am happy to see it go.

[Araneidae]

@coretl, I also have no ties to 3.6 ... except I would point out that it is the stock version of Python installed on RHEL8, so dropping it may have consequences for users on that platform.

[mdavidsaver]

... stock version of Python installed on RHEL8 ...

This is a fair point. I can think of two possibilities in addition to drop support for 3.6 or a full backport of asyncio.run.

State that the asyncio part is only supported with py >= 3.7. (fyi. driven by removal of loop= P4P only supports asyncio with py >= 3.6, but supports thread/cothread with py 2.7 and >= 3.4)

Continue the current behavior with <= 3.6

[coretl]

My preference is to drop 3.6, bump the major release number, and see if anyone complains... Any objections?

[AlexanderWells-diamond]

No objections here as long as its clear in the Changelog.

[mdavidsaver]

Is there anything I can do to move this PR forward?

[coretl]

Is there anything I can do to move this PR forward?

If you can drop 3.6 from the GitHub actions tests matrix, then add a test with your usecase (checking a spawned subprocess running as a Task exits cleanly outside context manager) then I'd be happy to merge this.

@AlexanderWells-diamond @Araneidae any thoughts?

[Araneidae]

I have learnt that there is a python3.11 rpm available on our RHEL8 channel and so presumably potentially available on all administered RHEL8 systems, so I agree that 3.6 can now go.

[AlexanderWells-diamond]

This PR will need some tidying up before we can merge it (remove 3.6, documentation, changelog etc.); I'll do that in the near future.

[AlexanderWells-diamond]

I've updated this PR to remove Python 3.6 support and added a Changelog entry.

[AlexanderWells-diamond]

With my addition of a test for the Context Manager, I'm finished with my changes. @Araneidae and/or @coretl please merge when you are happy with it.

[mdavidsaver]

@AlexanderWells-diamond Thank you for taking the time to finish this up!