Use AllowDiv when doing REPL paren matching.

Summary:
If we use `AllowRegExp` as the grammar context, then paren matching
breaks if we use `/` as division inside parens.

Always use `AllowDiv` instead. Note that this is not perfect,
because it doesn't work if there are unbalanced parens inside
an actual RegExp.

Reviewed By: tmikov

Differential Revision: D21434307

fbshipit-source-id: c3258bbe73d815349985f97629b54d9dc14e007f

test/repl/regress-div.js

@@ -0,0 +1,12 @@
+// Remove this once the REPL can handle block comments.
+// @lint-ignore-every LICENSELINT
+
+// Copyright (c) Facebook, Inc. and its affiliates.
+//
+// This source code is licensed under the MIT license found in the
+// LICENSE file in the root directory of this source tree.
+
+// RUN: cat %s | %hermes -prompt="" -prompt2="" | %FileCheck --match-full-lines %s
+
+(1 / 2);
+// CHECK: 0.5

tools/hermes/repl.cpp

@@ -193,7 +193,11 @@ static bool needsAnotherLine(llvm::StringRef input) {
     }
   };
 
-  while (const parser::Token *token = lexer.advance()) {
+  // Use AllowDiv to not skip parens when using division.
+  // TODO: Find a way to use the correct GrammarContext and make this work
+  // for all REPL inputs.
+  while (const parser::Token *token =
+             lexer.advance(parser::JSLexer::GrammarContext::AllowDiv)) {
     if (token->getKind() == parser::TokenKind::eof) {
       break;
     }