FIX: kernprof -m: presence of the executed module as sys.modules['__main__']

[TTsangSC]

Motivation

kernprof -m as it is now fails to address edge cases where the executed module is expected to be found at sys.modules['__main__']. One such cases is for the @enum.global_enum decorator, which searches sys.modules for the module whither to insert the enum instances. This causes use-cases like kernprof -m calendar to fail. Notably, a similar failure pattern is seen with even Python stdlib modules.

Changes

The key point of failure lies with how kernprof uses runpy.run_module() without the optional alter_sys argument. Since it defaults to false, the executed module is not inserted into sys.modules, and thus the aforementioned use-case fails. Hence, this PR implements the following changes:

• CHANGELOG.rst:
  Added entry
• kernprof.py::main():
  Now calling runpy.run_module() with alter_sys = True to maintain compatibility with code expecting the executed module to be found at sys.modules['__main__']
• line_profiler/autoprofile/autoprofile.py::run():
  Updated implementation to do the equivalent of runpy.run_module(alter_sys=True)
• tests/test_kernprof.py:
  • test_kernprof_m_parsing():
    • Updated implementation to expect the correct sys.argv[0] (real path to the module file)
    • Replaced tempfile.mkdtemp() with tempfile.TemporaryDirectory
  • test_kernprof_m_sys_modules():
    New test for compatibility with tools (e.g. @enum.global_enum) expecting the executed module to be found at sys.modules['__main__']

[TTsangSC]

We may want to give priority to this (over the other PRs) given that it's a fix instead of a new feature...

@Erotemic

[codecov]

Codecov Report

Attention: Patch coverage is 92.85714% with 2 lines in your changes missing coverage. Please review.

Project coverage is 63.63%. Comparing base (7f25e70) to head (53a3846).
Report is 8 commits behind head on main.

Files with missing lines	Patch %	Lines
line_profiler/autoprofile/autoprofile.py	92.85%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #339      +/-   ##
==========================================
+ Coverage   62.30%   63.63%   +1.33%     
==========================================
  Files          13       13              
  Lines        1008     1034      +26     
  Branches      225      229       +4     
==========================================
+ Hits          628      658      +30     
+ Misses        316      315       -1     
+ Partials       64       61       -3     

Files with missing lines	Coverage Δ
line_profiler/autoprofile/autoprofile.py	95.00% <92.85%> (-5.00%)	⬇️

... and 2 files with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 155c4c2...53a3846. Read the comment docs.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Erotemic]

Overall this looks great, and I had to go out of my way to find things to comment on. The only thing I think should be changed is the assert statement (unless I'm missing something and we really expect that to never happen, but it looks like a user-typo could cause that case to be hit).

[TTsangSC]

Thanks for the review! Just updated the PR.

[TTsangSC]

Weird, the pipeline isn't running after I pushed the changes... did we recently change workflow-triggering policies?

[Erotemic]

I haven't changed anything. Not sure why the CI isn't running. There isn't anything about a pending workflow that needs approval. It just isn't running and I don't see an option to force it to run.

EDIT: I added a dummy commit, and it seems to run fine now. Not sure what was up. I'm unsatisfied with this hack, but at the same time I don't want to spend effort looking into it. Will keep an eye out if it happens again.

[TTsangSC]

Cheers!