[regexp] Use SameValue in @@search as specced

test/mjsunit/regexp.js

 // Copyright 2012 the V8 project authors. All rights reserved.
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions are
 // met:
 //
 //     * Redistributions of source code must retain the above copyright
 //       notice, this list of conditions and the following disclaimer.
 //     * Redistributions in binary form must reproduce the above
 //       copyright notice, this list of conditions and the following
 //       disclaimer in the documentation and/or other materials provided
@@ skipping 721 matching lines @@
 assertEquals(["\u{daff}", "\u{e000}"], "\u{daff}\u{e000}".split(/[a-z]{0,1}/u));
 assertEquals(["\u{daff}\u{dfff}"], "\u{daff}\u{dfff}".split(/[a-z]{0,1}/u));
 
 // Test that changing a property on RegExp.prototype results in us taking the
 // slow path, which executes RegExp.prototype.exec instead of our
 // RegExpExecStub.
 const RegExpPrototypeExec = RegExp.prototype.exec;
 RegExp.prototype.exec = function() { throw new Error(); }
 assertThrows(() => "abc".replace(/./, ""));
 RegExp.prototype.exec = RegExpPrototypeExec;
+
+// Test the code path in RE.proto[@@search] when previousLastIndex is a receiver
+// but can't be converted to a primitive. This exposed a crash in the
+// C++ implementation of @@search.
+
+var re = /./;
+re.lastIndex = { [Symbol.toPrimitive]: 42 };
+() => "abc".search(re);

[Dan Ehrenberg, 2016/11/02 12:03:45]

Does this test actually run anything? Did you mean to have an assertThrows call
here?

[jgruber, 2016/11/07 07:56:18]

Good catch, thanks! This line was originally in an assertThrows, which tested a
specific code path in Object::Equals when lastIndex was not convertible to
primitive. 

However, assertThrows is no longer correct now that we use Object::SameValue
instead, which does not throw in this case.

Will fix in another CL.