buffer: properly retrieve binary length of needle

If the needle contains an extended latin-1 character then using
String::Utf8Length() will be too large and the search will return early.
Instead use String::Length() when encoding is BINARY.

PR-URL: nodejs#4803
Reviewed-By: James M Snell <jasnell@gmail.com>

src/node_buffer.cc

@@ -849,7 +849,9 @@ void IndexOfString(const FunctionCallbackInfo<Value>& args) {
   Local<String> needle = args[1].As<String>();
   const char* haystack = ts_obj_data;
   const size_t haystack_length = ts_obj_length;
-  const size_t needle_length = needle->Utf8Length();
+  // Extended latin-1 characters are 2 bytes in Utf8.
+  const size_t needle_length =
+      enc == BINARY ? needle->Length() : needle->Utf8Length();
 
 
   if (needle_length == 0 || haystack_length == 0) {

test/parallel/test-buffer-indexof.js

@@ -109,6 +109,15 @@ assert.equal(
 assert.equal(
     Buffer(b.toString('binary'), 'binary')
     .indexOf(Buffer('d', 'binary'), 0, 'binary'), 3);
+assert.equal(
+    Buffer('aa\u00e8aa', 'binary')
+    .indexOf('\u00e8', 'binary'), 2);
+assert.equal(
+    Buffer('\u00e8', 'binary')
+    .indexOf('\u00e8', 'binary'), 0);
+assert.equal(
+    Buffer('\u00e8', 'binary')
+    .indexOf(Buffer('\u00e8', 'binary'), 'binary'), 0);
 
 
 // test optional offset with passed encoding