Make the mypy-based tests faster!

[A5rocks]

This makes some of the tests slower cause it stops using an (implicit) cache in CWD. This makes them generally faster, however.

Fixes #2662

Edit: huh, did not realize I pressed the "assignees" button!

[codecov]

Codecov Report

Merging #2664 (1f75b1f) into master (493c915) will decrease coverage by 0.07%.
The diff coverage is 93.68%.

❗ Current head 1f75b1f differs from pull request most recent head 7be61ff. Consider uploading reports for the commit 7be61ff to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2664      +/-   ##
==========================================
- Coverage   98.15%   98.09%   -0.07%     
==========================================
  Files         118      118              
  Lines       16480    16536      +56     
  Branches     2983     3000      +17     
==========================================
+ Hits        16176    16221      +45     
- Misses        249      256       +7     
- Partials       55       59       +4     

Impacted Files	Coverage Δ
trio/_tests/test_exports.py	95.00% <93.68%> (-5.00%)	⬇️

[A5rocks]

Timeouts are annoying, I think fixing those (probably just making mypy run once without pytest timeout and share that cache whereever) should be another PR however. NVM I need to fix that to merge this...

Edit: also I see this ended up being slower by 2 minutes... huh??

[A5rocks]

In case it isn't obvious, this changed from sys.version_info >= (3, 12) to sys.version_info >= (3, 11) cause of a backport, probably.

[jakkdl]

We could also set up a shared cache dir across the jobs - I think we can assume that mypy will generate the same cache on different OS's, and could then also use that in check.sh.

Not sure if https://docs.github.com/en/actions/using-workflows/caching-dependencies-to-speed-up-workflows or https://docs.github.com/en/actions/using-workflows/storing-workflow-data-as-artifacts would be the way to go.

[jakkdl]

Timeouts are annoying, I think fixing those (probably just making mypy run once without pytest timeout and share that cache whereever) should be another PR however. NVM I need to fix that to merge this...

Edit: also I see this ended up being slower by 2 minutes... huh??

Is it still slower? Maybe mypy's cache parsing is sufficiently optimized that (only) sidestepping it doesn't actually improve it.

[A5rocks]

Timeouts are annoying, I think fixing those (probably just making mypy run once without pytest timeout and share that cache whereever) should be another PR however. NVM I need to fix that to merge this...
Edit: also I see this ended up being slower by 2 minutes... huh??

Is it still slower? Maybe mypy's cache parsing is sufficiently optimized that (only) sidestepping it doesn't actually improve it.

Nope, that comment's out of date. It's like 1/3rd faster now. (master takes ~2 minutes, this takes ~1:20)

We could also set up a shared cache dir across the jobs - I think we can assume that mypy will generate the same cache on different OS's, and could then also use that in check.sh.

Not sure if https://docs.github.com/en/actions/using-workflows/caching-dependencies-to-speed-up-workflows or https://docs.github.com/en/actions/using-workflows/storing-workflow-data-as-artifacts would be the way to go.

Eh, not too keen on this. Mypy's cache can lead to issues and I don't think an extra 10 seconds or so in the tests is worth the wackiness it leads to.

[jakkdl]

2 small changes, feel free to merge after addressing them!

[A5rocks]

Rewrote history a bit to make the merge look neater, feel free to merge via merge commit (do the honors!)!