Allow range-to-inclusive macro patterns

[chenyukang]

Fixes #120737

[rustbot]

r? @folkertdev

rustbot has assigned @folkertdev.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 75 candidates
• Random selection from 20 candidates

[folkertdev]

r? @nnethercote

Looks fine but it's macro parsing to there might be something subtle I'm missing

[nnethercote]

I think the "only used for diagnostics" is not true; can_begin_pattern is used in both nonterminal_may_begin_with and parse_pat_with_range_pat which affect code parsing.

View changes since the review

[chenyukang]

this was added only for diagnostic.

the parse_pat_with_range_pat invocation is also emitting error:

rust/compiler/rustc_parse/src/parser/pat.rs

Lines 734 to 744 in bcd9ae8

	if self.token.is_keyword(kw::Let)
	&& self.look_ahead(1, |tok| {
	tok.can_begin_pattern(token::NtPatKind::PatParam { inferred: false })
	})
	{
	self.bump();
	// Trim extra space after the `let`
	let span = lo.with_hi(self.token.span.lo());
	self.dcx().emit_err(RemoveLet { span: lo, suggestion: span });
	lo = self.token.span;
	}

the nonterminal_may_begin_with is indeed expand its usage, with a comment here:
https://github.com/rust-lang/rust/pull/120221/changes#r1669391427
and
#120221 (comment)

[nnethercote]

The change you've made is the obvious fix, but I'm uncertain about whether we can just accept the PR, or if we need some extra verification (e.g. crater run, or lang team involvement). @petrochenkov, what do you think?

[petrochenkov]

Any changes to fn can_begin_* functions can cause observable changes in macro matching, from what I remember.
I'd start with constructing an example showing how it changes in this specific case.

I probably wouldn't bother running crater for this specific change, unless it's rolled up with something else or as a part of beta run.

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[chenyukang]

Any changes to fn can_begin_* functions can cause observable changes in macro matching, from what I remember. I'd start with constructing an example showing how it changes in this specific case.

I probably wouldn't bother running crater for this specific change, unless it's rolled up with something else or as a part of beta run.

yes, this seems is a break change, here is an example where the selected macro arm changes:

macro_rules! select {
    ($p:pat) => { "pattern" };
    ($($tokens:tt)*) => { "fallback" };
}

fn main() {
    println!("{}", select!(..=0));
}

currently this prints "fallback". With this change, it would print "pattern", since the first arm would accept ..=0.

but the Reference already specifies that pat and pat_param accept pattern grammar that includes ..=N?

seems this is a difference between our current code with reference.