cc @wide-midnight-78598

Specifically, it appears to have an issue with

concurrently()

awaiting multiple calls-by-name

I see https://github.com/pantsbuild/pants/pull/21144 by @worried-painter-31382

The issue seems to be that here the elements we call the

get()

helper on are the raw coroutines, not the

Call

wrappers.

Indeed, this is

tuple[Get | Coroutine, ...]

and not

tuple[Get | Call, ...]

Module patching is "the way" to mock functions in python, after all. Sure, the tests would need to be async and the mocks has to return awaitables, but writing utilities for that is like 2 lines of code and

pytest-asyncio

works fine

concurrently

complicates that option a little bit, but I think maybe deprecating

run_rule_with_mocks

and providing a facility to mock

concurrently

, possibly a patching context manager that just causes

concurrently

to return the input arguments ~ as is would be much more intuitive in a unit test context.

Argument being that

MockGet

mirrors

Get

, not

call_by_name

.

It’s a bit more complicated, since the named thing you naively patch is actually the

Call

wrapper, not it’s underlying

.rule.func

, and that thing is a field in a frozen dataclass so you can’t naively patch on that either…

But I think I will be able to get this to work with some Python hammers