fix: respect trailing semicolon setting for last statements

[shekhirin]

Problem

#5797 fixed the trailing_semicolon rule, allowing it only for the last expression in the block. The side-effect of this change is that such snippet

fn main() {
    println!("{}", greet());
}

fn greet() -> String {
    return "Hello, b!".to_string();
}

will not get formatted into

fn main() {
    println!("{}", greet());
}

fn greet() -> String {
    return "Hello, b!".to_string()
}

(notice the ; in the greet function), even though return is the last expression here. It's happening because the implementation of is_last_expr() additionally checks for the type of expression for some reason unknown to me

rustfmt/src/stmt.rs

Lines 74 to 88 in 46e5f14

	fn is_last_expr(&self) -> bool {
	if !self.is_last {
	return false;
	}
	match self.as_ast_node().kind {
	ast::StmtKind::Expr(ref expr) => match expr.kind {
	ast::ExprKind::Ret(..) | ast::ExprKind::Continue(..) | ast::ExprKind::Break(..) => {
	false
	}
	_ => true,
	},
	_ => false,
	}
	}

Solution

This PR fixes that by checking only the is_last field of the Stmt struct when deciding whether to emit the semicolon.

fn main() {
    println!("{}", greet());
}

fn greet() -> String {
    return "Hello, b!".to_string();
}

fn foo() {}
fn main() {
    return;
    foo()
}

➜  rustfmt git:(trailing-semicolon-last-stmt) cargo run --bin rustfmt -- --check -- ~/Projects/semicolon/src/main.rs
   Compiling rustfmt-nightly v1.7.0 (/Users/shekhirin/Projects/rustfmt)
    Finished dev [unoptimized + debuginfo] target(s) in 0.61s
     Running `target/debug/rustfmt --check -- /Users/shekhirin/Projects/semicolon/src/main.rs`
Diff in /Users/shekhirin/Projects/semicolon/src/main.rs:3:
 }
 
 fn greet() -> String {
-    return "Hello, b!".to_string();
+    return "Hello, b!".to_string()
 }
 
 fn foo() {}

As you can see, the example from the original PR is still not getting formatted, while the trailing comma for the last return statement gets removed.

[ytmimi]

Will probably be a little while before I'm able to review this, but I wanted to note that there's a semi related PR (#6149) that adds a new option to trailing_semicolon. It feels like this would be complimentary to that work, but I'm not 100% sure.

[rustbot]

☔ The latest upstream changes (possibly #6823) made this pull request unmergeable. Please resolve the merge conflicts.