Release 0.12.0

[shamrin]

Release notes screenshot

TODO

• [x] get #96 merged
• [x] enable xfailed test (fixed by #96)
• [x] rename test_asyncio_trio_adapted_no_loop to something else
• [x] get review
• [x] get #99 merged
• [x] get #96 and #99 news fragments into history.rst
• [x] update the date in history.rst
• [x] investigate #103
• [x] tag with vVERSION, push tag
• [x] PyPI build
• [x] update version number in the same pull request: add +dev tag to the end
• [x] merge this PR
• [x] announce on gitter

Fixes #93

[smurfix]

The test error is from Python 3.10 which dropped a heap of loop=loop parameters from Future, Event et al.

I'm building a minimal patch; more extensive de-loop-argument-ization (in trio-asyncioproper) makes sense but requires deprecation warnings.

[codecov[bot]]

Codecov Report

Merging #94 (a18c2b4) into master (b93c320) will increase coverage by 1.16%. The diff coverage is 87.50%.

@@            Coverage Diff             @@
##           master      #94      +/-   ##
==========================================
+ Coverage   80.64%   81.81%   +1.16%     
==========================================
  Files          11       11              
  Lines        1240     1193      -47     
  Branches      172      172              
==========================================
- Hits         1000      976      -24     
+ Misses        170      154      -16     
+ Partials       70       63       -7     

Impacted Files	Coverage Δ
trio_asyncio/__init__.py	100.00% <ø> (ø)
trio_asyncio/_loop.py	79.32% <50.00%> (+0.63%)	:arrow_up:
trio_asyncio/_adapter.py	81.19% <100.00%> (+1.03%)	:arrow_up:
trio_asyncio/_base.py	84.14% <100.00%> (-0.64%)	:arrow_down:
trio_asyncio/_util.py	84.09% <100.00%> (ø)
trio_asyncio/_version.py	100.00% <100.00%> (ø)
trio_asyncio/_deprecate.py	100.00% <0.00%> (+19.44%)	:arrow_up:

[shamrin]

@smurfix Thank you! I considered doing this change myself, but I thought it would make trio-asyncio Python 3.8+ only. Test results are proving me wrong :)

[shamrin]

Code coverage got lower. Most of it is due to _deprecate module. We use much less of it now, due to removed deprecations. Options (ordered by increasing difficulty):

• copy test_deprecate from Trio and adapt it (easiest)
• remove unused functionality from _deprecate module
• extract _deprecate to its own package and use it in both trio and trio-asyncio

Update: I did port test_deprecate from Trio and codecov is finally happy. 👇

[shamrin]

@oremanj:

I just uploaded #96 which appears to fix it. Probably worth incorporating into the release?

Awesome, thank you! Yes, let's postpone the release until we get #96 merged.

(In the mean time, I've made the fixes you suggested.)

[oremanj]

This LGTM. I merged #96 so once you incorporate that (and #99 maybe?) I think it's ready.

[oremanj]

Interesting that this PR hasn't had any of the problems that the Mac builds have been having on other PRs. Those all seem to have failed around test_cancel_sleep_depr in test_misc.py, which is now gone. But I can't for the life of me figure out why test_cancel_sleep_depr would hang if the equivalent test_cancel_sleep works fine. Gremlins!

[shamrin]

@oremanj Now it did stuck at test_cancel_sleep. And I did find a way to replicate the issue locally. Issue coming soon!

Update: here is the issue - #103. Should we fix this deadlock before the release? Opinions?

[shamrin]

"Fixed" the #103 deadlock in test_cancel_sleep by removing unnecessary raise Exception. The test would now fail with AssertionError rather than deadlock. I've also fiddled with timeout length to lower failure probability: 970d81c.

I think the underlying cause of deadlock is #104. No need to fix it for this release, because the problem was in v0.11.0 as well.