expand: Leave traces when expanding cfg attributes

[petrochenkov]

This is the same as #138515, but for cfg(true) instead of cfg_attr.

The difference is that cfg(true)s already left "traces" after themselves - the cfg attributes themselves, with expanded_inert_attrs set to true, with full tokens, available to proc macros.
This is not a reasonably expected behavior, but it could not be removed without a replacement, because a major rustdoc feature and a number of clippy lints rely on it. This PR implements a replacement.

This needs a crater run, because it changes observable behavior (in an intended way) - proc macros can no longer see expanded cfg(true) attributes.

(Some minor unnecessary special casing for sym::cfg_attr is also removed in this PR.)

r? @nnethercote

[rustbot]

Some changes occurred in src/tools/clippy

cc @rust-lang/clippy

Some changes occurred in compiler/rustc_passes/src/check_attr.rs

cc @jdonszelmann

[petrochenkov]

@bors try

[bors]

⌛ Trying commit 750814d with merge 411f3ca...

[bors]

☀️ Try build successful - checks-actions
Build commit: 411f3ca (411f3ca0d8086d5a27e2aaa79b9ff41f53ab2f51)

[petrochenkov]

@craterbot check

[craterbot]

👌 Experiment pr-138844 created and queued.
🤖 Automatically detected try build 411f3ca
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🚧 Experiment pr-138844 is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🎉 Experiment pr-138844 is completed!
📊 6 regressed and 7 fixed (602721 total)
📰 Open the full report.

⚠️ If you notice any spurious failure please add them to the denylist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[flip1995]

Starting a thread on a random file: Just noticed those changes and had a question:

In Clippy we have this utils function:

rust/src/tools/clippy/clippy_utils/src/attrs.rs

Lines 177 to 198 in b95aac6

	/// Checks if the given span contains a `#[cfg(..)]` attribute
	pub fn span_contains_cfg(cx: &LateContext<'_>, s: Span) -> bool {
	s.check_source_text(cx, |src| {
	let mut iter = tokenize_with_text(src);
	// Search for the token sequence [`#`, `[`, `cfg`]
	while iter.any(|(t, ..)| matches!(t, TokenKind::Pound)) {
	let mut iter = iter.by_ref().skip_while(|(t, ..)| {
	matches!(
	t,
	TokenKind::Whitespace | TokenKind::LineComment { .. } | TokenKind::BlockComment { .. }
	)
	});
	if matches!(iter.next(), Some((TokenKind::OpenBracket, ..)))
	&& matches!(iter.next(), Some((TokenKind::Ident, "cfg", _)))
	{
	return true;
	}
	}
	false
	})
	}

This function is there to check after-the-fact if there was a cfg involved in the code that we're currently checking. For example when we're looking at a match:

match x {
    #[cfg(foo)]
    optionalArm => {},
    alwaysArm => {},
}

And then gating linting on it.

IIUC this can help us getting rid of this utils function? If so, I'd open a issue on Clippy for it. Definitely not for this PR.

[petrochenkov]

cfg(true) already stays in code after expansion, and some other logic in clippy checks for it, yet this function still exists.

Maybe it takes a span of some larger node and expects to see #[cfg(false)]s in it as well? In that case this PR won't help, configured out nodes still disappear without a trace.