Less maybe_whole_expr

[nnethercote]

r? @petrochenkov

[petrochenkov]

e is not actually checked for being a literal here.
What happens if non-literal NtExpr ends up being passed to this function?

[nnethercote]

Good question. I don't know. But the existing code (which uses maybe_whole_expr) also doesn't check if e is a literal. I will try adding a is-literal check and see if anything fails.

[nnethercote]

Adding a literal check causes various test failures:

    [ui] tests/ui/issues/issue-43250.rs
    [ui] tests/ui/macros/macro-pat-neg-lit.rs
    [ui] tests/ui/macros/vec-macro-in-pattern.rs
    [ui] tests/ui/match/expr_before_ident_pat.rs
    [ui] tests/ui/parser/issues/issue-65122-mac-invoc-in-mut-patterns.rs
    [ui] tests/ui/parser/issues/issue-70050-ntliteral-accepts-negated-lit.rs
    [ui] tests/ui/pattern/issue-92074-macro-ice.rs
    [ui] tests/ui/pattern/patkind-litrange-no-expr.rs
    [ui] tests/ui/proc-macro/expand-expr.rs

Some of these are just changes in error messages, e.g.:

---- [ui] tests/ui/issues/issue-43250.rs stdout ----
diff of stderr:

-	error: arbitrary expressions aren't allowed in patterns
-	  --> $DIR/issue-43250.rs:9:8
+	error: expected pattern, found `y`
+	  --> $DIR/issue-43250.rs:6:17
3	   |
+	LL |             let $a = 0;
+	   |                 ^^ expected pattern
+	...
4	LL |     m!(y);
-	   |        ^
+	   |     ----- in this macro invocation
+	   |
+	   = note: this error originates in the macro `m` (in Nightly builds, run with -Z macro-backtrace for more info)
6	
-	error: arbitrary expressions aren't allowed in patterns
-	  --> $DIR/issue-43250.rs:11:8
+	error: expected pattern, found `C`
+	  --> $DIR/issue-43250.rs:6:17
9	   |
+	LL |             let $a = 0;
+	   |                 ^^ expected pattern
+	...
10	LL |     m!(C);
-	   |        ^
+	   |     ----- in this macro invocation
+	   |
+	   = note: this error originates in the macro `m` (in Nightly builds, run with -Z macro-backtrace for more info)
12	
13	error: aborting due to 2 previous errors
14	

That one now fails while parsing, instead of failing during lowering.

But some check-pass tests now fail to compile, e.g.:

error: unexpected token: `-2`
  --> fake-test-src-base/parser/issues/issue-70050-ntliteral-accepts-negated-lit.rs:5:14
   |
LL |         bar!($a)
   |              ^^
...
LL |     ($b:literal) => {};
   |      ---------- while parsing argument for this `literal` macro fragment
...
LL |     foo!(-2);
   |     -------- in this macro invocation
   |
   = note: this error originates in the macro `foo` (in Nightly builds, run with -Z macro-backtrace for more info)

So I don't think a literal check is appropriate here.

[petrochenkov]

The function is called parse_literal_maybe_minus so it's not surprising that it accepts -lit in addition to just lit.
I meant checking for anything except lit and -lit (and maybe ExprKind::Err).

[petrochenkov]

We need at least a test executing these case, if not a check.
Perhaps they are reported as errors at some later point, then an extra check won't be necessary.

[nnethercote]

I have updated the code, with some more minor refactorings. But I didn't add the is-literal check because of the test failures mentioned above.

[bors]

☔ The latest upstream changes (presumably #107679) made this pull request unmergeable. Please resolve the merge conflicts.

[nnethercote]

I don't think this is worth the effort, but I opened #109415 for the refactoring of handle_missing_lit.