Type completions in tuple structs/variants are broken by attributes

[jonas-schievink]

struct Newtype(#[serde] Old$0);

struct Oldtype;

This does not complete Oldtype (or anything else).

[yue4u]

I dug into this a bit and found it was disabled intentionally at 11115eb in #9301. Since is_path_disallowed been marked as "shouldn't exist" in 25d133e#diff-563c0914dd89f1c5e6ca3ae5eb22105f0bcd42c15ae26580a4e7a25dc10d9a96R296 I guess this issue will be resolved as a part of #12144?

[Veykril]

I just looked into this as well 😅, and yes that is the problem. We can actually remove that attribute rule from is_path_disallowed as our completions seem to be good enough to already handle the unwanted cases covered by it.

[Veykril]

I'll add the functions that shouldn't exist to that tracking issue as tasks actually.