fix wasm build and tests

[samuelcolvin]

try using pytest-speed as a substitute for pytest-benchmark for wasm builds.

[codecov]

Codecov Report

Merging #148 (3594bbd) into main (158ab9a) will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##             main     #148   +/-   ##
=======================================
  Coverage   97.71%   97.71%           
=======================================
  Files          44       44           
  Lines        4422     4422           
  Branches       31       31           
=======================================
  Hits         4321     4321           
  Misses        101      101           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 158ab9a...3594bbd. Read the comment docs.

[hoodmane]

Note that pytest-benchmark works on Pyodide's main branch, though I don't know what the differences are between pytest-benchmark and pytest-speed.

OverflowError: timestamp out of range for platform time_t

There are also errors with datetime on PyO3:
https://github.com/PyO3/pyo3/pull/2436/files#diff-0475f1ee5a37fa84fc40bdb6aa3e0e463ebbd19d67fa60b5872244497480dbd3
I assumed that it was not a real bug and is due to some dumb configuration mistake but seeing it here makes it more likely to be real. Someone will have to investigate.

[samuelcolvin]

Interesting, thanks.

I've been working on pytest-speed today to fix some frustrations with pytest-benchmark, it's backwards compatible with pytest-benchmark in this codebase. I thought the problem was with a dependency of pytest-benchmark which didn't distribute wheels? Either way, over time I'll replace pytest-benchmark complete in time.

[hoodmane]

The problem is that py-cpuinfo doesn't support wasm. It would be an easy fix but py-cpuinfo doesn't look maintained.

[samuelcolvin]

The whole of pytest-benchmark looks pretty unloved I'm afraid.

Quite about from the hard to read output, the actual measurements seem to vary a lot, hence why I went down the rabbit whole of starting anotherpackage.