timeout coverage/pytest with sigint to help debug CI timeout issues

[jakkdl]

didn't find any dead simple ways of doing this within pytest. I hope this works on other platforms

[codecov]

Codecov Report

Merging #2827 (ec87c56) into master (315a0e8) will decrease coverage by 0.07%.
The diff coverage is n/a.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2827      +/-   ##
==========================================
- Coverage   99.18%   99.12%   -0.07%     
==========================================
  Files         115      115              
  Lines       17551    17551              
  Branches     3149     3149              
==========================================
- Hits        17408    17397      -11     
- Misses         99      109      +10     
- Partials       44       45       +1     

see 1 file with indirect coverage changes

[jakkdl]

wait hmm, the timeouts happening on Ubuntu (pypy-3.9) and Ubuntu (pypy-3.10) are some very different type of timeout than the CI being mad ... I think??
It's at least talking about some 1min timeout being triggered

[jakkdl]

Okay, actually digging into the traces I now realize this isn't about the CI killing at all - this seems to be Trio itself giving up in the main event loop. And it's not any particular test triggering it.

Manually going back and checking runs, the oldest I can find is https://github.com/python-trio/trio/actions/runs/6442036631/job/17492440655 .... in a PR that is modifying thread behaviour. So it's possible that #2392 has some subtle bugs, and should maybe be reverted for now?

[A5rocks]

TBH this PR isn't needed for an immediate issue anymore and does complicate CI a bit. Do you think it's still worth pursuing?

[jakkdl]

not especially. I do like the extra refactoring I did of pulling apart the previously huge and tricky-to-parse line though.

[jakkdl]

although: https://github.com/python-trio/trio/actions/runs/6641043117/job/18042685943?pr=2820

weirdly enough it displays as having passed despite being canceled???

[A5rocks]

Yeah I'm not sure what's up with that CI pass :(

I'm going to hope it's gone away. And we've cleaned up the line a bit more. Feel free to open a new PR (or reopen this and fix conflicts) if you still think it would benefit from more refactoring!