rustdoc::formats::Cache does not contain enough data to unwrap type aliases

[lolbinarycat]

rustdoc has a lot of issues handling type aliases, and I think a large part of that is due to the fact that once the cleaned ast is constructed, type aliases are essentially opaque, and code cannot see what type they point to.

clean::TypeAlias does exist, but there's no way to get that from a Path item, which is what shows up in type args.

I encountered this when working on #138574

[lolbinarycat]

[IRLO] yet another feature request blocked on this

[lolbinarycat]

I see 3 main ways this could be implemented:

1. hold onto the TyCtxt for type alias resolution (kinda breaks the current model of rustdoc where rendering only depends on the cleaned AST)
2. collect a map of all type aliases by hooking rustdoc::clean::inline::build_type_alias (simple, but may have significant memory overhead)
3. collect a map of only the type aliases that are needed for rendering (eg. only type aliases that appear in the generic parameters of a deref Target type) (less memory usage, but significantly more complicated, and likely uses more cpu cycles)

[notriddle]

https://rust-lang.zulipchat.com/#narrow/channel/393423-t-rustdoc.2Fmeetings/topic/2025-04-14/near/512157591

In the meeting, it seemed like this issue needs more analysis. It seems to be related to Lazy Type Aliases, but that's not sure.

Is there a minimal reproduction test case for this?

[lolbinarycat]

@notriddle sorry, i don't quite understand. this is more of a meta/tracking issue, how would i provide a MCVE? did you mean to comment on #138614?