Show name of type variable in "Cannot infer type argument" message

[brianschubert]

Fixes #19289

The type argument index currently shown can be wrong if other type arguments were substituted during an earlier pass.

Given:

def foo[T1, T2](
    a: T1,
    b: T2,
    c: Callable[[T2], T2],
) -> tuple[T1, T2]: ...

def bar(y: float) -> float: ...

reveal_type(foo(1, None, bar))  # Expect T1=int, T2=<failed>

Before:

main.py:9: error: Cannot infer type argument 1 of "foo"  [misc]
main.py:9: note: Revealed type is "tuple[builtins.int, Any]"

After:

main.py:9: error: Cannot infer type argument to type parameter "T2" of "foo"  [misc]
main.py:9: note: Revealed type is "tuple[builtins.int, Any]"

[brianschubert]

primer shard 0 died due to hitting the 60 minute timeout :(

[sterliakov]

LG! Only a minor optional nit:

[brianschubert]

I'm a little confused how a new test case in master is leaking into this PR's CI: https://github.com/python/mypy/actions/runs/15734614442/job/44343883908?pr=19290#step:9:210

The failing case was added in #19279. That commit isn't in this branch's history and the failing case does not exist locally.

Trivial to fix, but still weird.

[JelleZijlstra]

"type argument to type parameter" feels unnecessarily verbose. I think this is clear enough:

Suggested change

	f"Cannot infer type argument to type parameter {format_type(tv, self.options)} of {callee_name}",
	f"Cannot infer type argument {format_type(tv, self.options)} of {callee_name}",

[brianschubert]

Yeah, I agree with this being verbose. Cannot infer type argument "T" originally sounded off to me since T isn't a type argument, it's the type parameter whose argument mypy is trying to infer.

Maybe Cannot infer value of type parameter "T" of "f"? I'm also fine with the suggested wording if that's your preference.

[sterliakov]

how a new test case in master is leaking into this PR's CI

actions/checkout checks out the merge base, not the PR HEAD. That's why actions do not run if PR has merge conflicts, for example. So tests are never run against your head revision, but rather against the result of merging your branch into the destination branch as of now.

[brianschubert]

TIL, thank you!

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

prefect (https://github.com/PrefectHQ/prefect)
- src/prefect/task_engine.py:648: error: Cannot infer type argument 2 of "update_for_task" of "ResultStore"  [misc]
+ src/prefect/task_engine.py:648: error: Cannot infer value of type parameter "R" of "update_for_task" of "ResultStore"  [misc]
- src/prefect/task_engine.py:1233: error: Cannot infer type argument 2 of "update_for_task" of "ResultStore"  [misc]
+ src/prefect/task_engine.py:1233: error: Cannot infer value of type parameter "R" of "update_for_task" of "ResultStore"  [misc]

werkzeug (https://github.com/pallets/werkzeug)
- src/werkzeug/datastructures/structures.py:711: error: Cannot infer type argument 2 of "setdefault" of "MutableMapping"  [misc]
+ src/werkzeug/datastructures/structures.py:711: error: Cannot infer value of type parameter "_VT" of "setdefault" of "MutableMapping"  [misc]