Rollup merge of #63459 - eddyb:issue-63430, r=petrochenkov

syntax: account for CVarArgs being in the argument list.

Fixes #63430 by testing for `1` argument (the `CVarArgs` itself) instead of `0`.

Note that the error has basically been impossible to trigger since the change that caused #63430, so perhaps we need an audit of untested errors.

Also, this check probably belongs in AST validation/HIR lowering, but I'd rather fix it in place for now.

r? @petrochenkov cc @dlrobertson

src/libsyntax/parse/parser.rs

@@ -5409,7 +5409,7 @@ impl<'a> Parser<'a> {
 
         let args: Vec<_> = args.into_iter().filter_map(|x| x).collect();
 
-        if c_variadic && args.is_empty() {
+        if c_variadic && args.len() <= 1 {
             self.span_err(sp,
                           "C-variadic function must be declared with at least one named argument");
         }

src/test/ui/c-variadic/variadic-ffi-no-fixed-args.rs

@@ -0,0 +1,6 @@
+extern {
+    fn foo(...);
+    //~^ ERROR C-variadic function must be declared with at least one named argument
+}
+
+fn main() {}

src/test/ui/c-variadic/variadic-ffi-no-fixed-args.stderr

@@ -0,0 +1,8 @@
+error: C-variadic function must be declared with at least one named argument
+  --> $DIR/variadic-ffi-no-fixed-args.rs:2:11
+   |
+LL |     fn foo(...);
+   |           ^
+
+error: aborting due to previous error
+