Isn't there a way from within a rule to call an async function that issues an

await Get

in its body? I'm getting this error when I do that:

E       native_engine.IntrinsicError: Get(Targets, UnparsedAddressInputs, UnparsedAddressInputs(values=('//runners/noop_runner',), relative_to='src/foobar', description_of_origin='entry_point_dependencies from src/foobar/test_something.py:tests', skip_invalid_addresses=False)) was not detected in your @rule body at rule compile time. Was the `Get` constructor called in a non async-function, or was it inside an async function defined after the @rule? Make sure the `Get` is defined before or inside the @rule body.

Here's the

Get

in the helper function that 2 rules call: https://github.com/pantsbuild/pants/pull/21062/files#diff-e1ddf55fc4b2b832db8a39ec88ffed8fdda413dde2620f56a5a18431882a6954R204-R211

I thought this used to be the use case for

@rule_helper

- but I thought that this should "just work" now...

Yeah. That's what I thought... I'm getting this error in a test, so I suppose it could be that I haven't registered all the rules I need to. But, the error clearly mentions compile time detection of the get.

Unless I'm misunderstanding, this should work? https://github.com/pantsbuild/pants/pull/17947

Oh. Yeah:

was it inside an async function defined after the @rule?

I'm reordering the rules now.

🥳 I ran into that a while back, and thought to myself “I’ll never forget this” - until I subsequently forgot about it

Related - looks like we can't call-by- name recursively? It gives that

was not detected in your @rule body at rule compile time

error.

Like, actually recursively? Do we recursively use GETs?

We should clarify this in the error message, that order of declaration matters.

Actually recursively. You can call a Get by type recursively inside the @rule that matches the type signature, but not by name.

You can call a Get by type recursively inside the @rule that matches the type signature,

Ah yeah Well, they'll run side-by-side, and plugins will have GETs, so it's an issue to flag. Probably not a huge issue in the grand scheme

Hmm, it seems like a huge issue?

Yeah, there's still a ton of sharp edges to go through, and manually fix things that we can't easily automate - e.g. cyclic deps

I'm debugging it ATM, will see what it's about