Skip to content
Commit c0cc19b5 authored by Richard Si's avatar Richard Si
Browse files

Delay worker count determination

os.cpu_count() can return None (sounds like a super arcane edge case
though) so the type annotation for the `workers` parameter of
`black.main` is wrong. This *could* technically cause a runtime
TypeError since it'd trip one of mypyc's runtime type checks so we
might as well fix it.

Reading the documentation (and cross-checking with the source code),
you are actually allowed to pass None as `max_workers` to
`concurrent.futures.ProcessPoolExecutor`. If it is None, the pool
initializer will simply call os.cpu_count() [^1] (defaulting to 1 if it
returns None [^2]). It'll even round down the worker count to a level
that's safe for Windows.

... so theoretically we don't even need to call os.cpu_count()
ourselves, but our Windows limit is 60 (unlike the stdlib's 61) and I'd
prefer not accidentally reintroducing a crash on machines with many,
many CPU cores.

[^1]: https://docs.python.org/3/library/concurrent.futures.html#concurrent.futures.ProcessPoolExecutor
[^2]: https://github.com/python/cpython/blob/a372a7d65320396d44e8beb976e3a6c382963d4e/Lib/concurrent/futures/process.py#L600
parent afed2c01
Loading
Loading
Loading
Loading
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Please register or to comment