Handle empty BigInt literals in the parser

Summary:
OSS-Fuzz found that empty octal/hex/binary BigInt literals trigger an
assert in our parser, since we don't correctly handle the case where
there are no digits after the 0o/0x/0b.

For example, `0bn` triggers this assert because after consuming the
trailing "n", we assume that some digits were consumed, which isn't
necessarily true.

Reviewed By: tmikov

Differential Revision: D37755199

fbshipit-source-id: 48805e0684ce4518a10e32ce17b02e22308444fa

lib/Parser/JSLexer.cpp

@@ -1461,13 +1461,13 @@ end:
 
     llvh::StringRef raw{rawStart, (size_t)(curCharPtr_ - rawStart)};
     if (ok && !real && (!legacyOctal || raw == "0n") && tmpStorage_ == "n") {
-      assert(curCharPtr_ > start + 1 && "there should be numbers here");
+      assert(curCharPtr_ > start && "Must consume at least the trailing n.");
+      llvh::ArrayRef<char> digits{start, curCharPtr_ - 1};
       // use parseIntWithRadix to validate the bigint literal's digits. The
       // converted value does not matter, only whether or not the string was
       // parsed correctly.
-      if (parseIntWithRadix</* AllowNumericSeparator */ true>(
-              llvh::ArrayRef<char>{start, (size_t)(curCharPtr_ - start - 1)},
-              radix)) {
+      if (digits.size() &&
+          parseIntWithRadix</* AllowNumericSeparator */ true>(digits, radix)) {
         // This is a BigInt.
         rawStorage_.clear();
         rawStorage_.append(raw);

test/Parser/literal.js

@@ -17,3 +17,6 @@
 
 0b2n
 // CHECK: invalid numeric literal
+
+0bn
+// CHECK: invalid numeric literal