Fix out-of-range access in unibrow::Utf8::CalculateValue.

test/unittests/unicode-unittest.cc

Index: test/unittests/unicode-unittest.cc
diff --git a/test/unittests/unicode-unittest.cc b/test/unittests/unicode-unittest.cc
new file mode 100644
index 0000000000000000000000000000000000000000..67edfb7331705c47a3716d32060942916c63a84b
--- /dev/null
+++ b/test/unittests/unicode-unittest.cc
@@ -0,0 +1,39 @@
+// Copyright 2016 the V8 project authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include <memory>
+#include <string>
+
+#include "src/unicode-decoder.h"
+#include "testing/gtest/include/gtest/gtest.h"
+
+namespace v8 {
+namespace internal {
+
+namespace {
+
+using Utf8Decoder = unibrow::Utf8Decoder<512>;
+
+void Decode(Utf8Decoder* decoder, const std::string& str) {
+  // Put the string in its own buffer on the heap to make sure that
+  // AddressSanitizer's heap-buffer-overflow logic can see what's going on.
+  std::unique_ptr<char[]> buffer(new char[str.length()]);
+  memcpy(buffer.get(), str.data(), str.length());
+  decoder->Reset(buffer.get(), str.length());
+}
+
+}  // namespace
+
+TEST(UnicodeTest, ReadOffEndOfUtf8String) {
+  Utf8Decoder decoder;
+
+  // Not enough continuation bytes before string ends.

[jbroman, 2016/11/22 00:25:04]

All of these cases produce AddressSanitizer failures if unittests is run with
is_asan=true.

[vogelheim, 2016/11/22 09:25:53]

Err... I take it they would with the original code, but no longer after this
fix, right?

[jbroman, 2016/11/22 12:40:44]

Yes, of course. :-)

+  Decode(&decoder, "\xE0");
+  Decode(&decoder, "\xED");
+  Decode(&decoder, "\xF0");
+  Decode(&decoder, "\xF4");
+}
+
+}  // namespace internal
+}  // namespace v8