impl Deref for EndianSlice appears in std docs

[Nemo157]

impl Deref for EndianSlice appears on the current nightly'sstd::ops::Deref documentation page, even though it's not part of the (public) standard libraries. This appears to be gimli::read::EndianSlice, so maybe related to #73441, cc @alexcrichton.

Noticed by @rijenkii (assuming they use the same handle here and discord, if not, sorry 😳)

[jyn514]

Relevant PRs (backtrace isn't really to blame here): #73771, #74534

[jyn514]

rust/src/librustdoc/clean/inline.rs

Lines 349 to 355 in e24a017

	if let Some(trait_did) = associated_trait {
	if let Some(stab) = cx.tcx.lookup_stability(trait_did.def_id) {
	if stab.level.is_unstable() {
	return;
	}
	}
	}

Maybe this should also skip the impl if there is no 'stability' attribute?

https://doc.rust-lang.org/nightly/nightly-rustc/rustc_middle/ty/query/queries/struct.lookup_stability.html, https://doc.rust-lang.org/nightly/nightly-rustc/rustc_attr/builtin/struct.Stability.html

[jyn514]

Maybe this should also skip the impl if there is no 'stability' attribute?

This is the default for everything not in the standard library. So it seems like the standard library needs special casing here. Maybe as simple as || cx.tcx.crate_name(LOCAL_CRATE) == "core" && stab.is_none()?

[bjorn3]

Doesn't gimli get everything marked as unstable behind the rustc_private feature gate through a rustc flag?

[tesuji]

MCVE: https://github.com/lzutao/rustdoc-traits-bugs
It doesn't require any external deps, just dummy crates included.
The outermost crate is a fake std, realcore is fake core, gimli is fake gimli.