add check for ConstKind::Value(_) to in_operand()

[mj10021]

Added check for valtree value to close #113012 which fixes the issue, although I am not sure if adding the check there is sound or not cc @oli-obk

[rustbot]

r? @davidtwco

(rustbot has picked a reviewer for you, use r? to override)

[mj10021]

should I add a copy the example from the issue into a test?

[oli-obk]

It's not wronger than the code path for ConstantKind::Val(..), though we should analyze at some point whether either of these are entirely correct here. Please remove the valtree fixme above now that valtrees are addressed.

[davidtwco]

r? @oli-obk

[mj10021]

It's not wronger than the code path for ConstantKind::Val(..), though we should analyze at some point whether either of these are entirely correct here. Please remove the valtree fixme above now that valtrees are addressed.

Happy to do any work on that if it would be helpful, although I probably would need some guidance

[oli-obk]

@bors r+ rollup

Happy to do any work on that if it would be helpful, although I probably would need some guidance

I believe the ConstValue::Val and ConstKind::Value cases can only happen for primitives, as they come from literals (1, "foo", ...). So what we could do is check that the type is a type that can't possibly violate any of the checks that we have here (&str, the inter and float types, &[u8], I think those are all that could show up here).

[bors]

📌 Commit 71362c7 has been approved by oli-obk

It is now in the queue for this repository.

[mj10021]

@bors r+ rollup

Happy to do any work on that if it would be helpful, although I probably would need some guidance

I believe the ConstValue::Val and ConstKind::Value cases can only happen for primitives, as they come from literals (1, "foo", ...). So what we could do is check that the type is a type that can't possibly violate any of the checks that we have here (&str, the inter and float types, &[u8], I think those are all that could show up here).

Ok great I can give that a go