Fix creating substring in string.replace(<global regexp>, <function>).

test/mjsunit/regexp-global.js

Index: test/mjsunit/regexp-global.js
diff --git a/test/mjsunit/regexp-global.js b/test/mjsunit/regexp-global.js
index fac5c6af03fd11dfcb3a4527d948c0e56b47315d..4b833d7c6c69a037878e8ae13109e38a69250810 100644
--- a/test/mjsunit/regexp-global.js
+++ b/test/mjsunit/regexp-global.js
@@ -125,3 +125,8 @@ str = str.replace(/(FOUR|TWO) \u817f (GOOD|BAD)/g,
                     return match.length - 7;
                   });
 assertEquals("4, 2!", str);
+
+// Test capture that is a real substring.

[Lasse Reichstein Nielsen, 2012/05/29 09:06:52]

Description isn't very descriptive :)
You are testing that a regexp that matches the empty string will match once at
the end of the string?

[Lasse Reichstein Nielsen, 2012/05/29 10:41:28]

Actually, you are testing a capture that *isn't* a proper substring.

+var str = "Beasts of England, beasts of Ireland";
+str = str.replace(/(.*)/g, function(match) { return '~'; });
+assertEquals("~~");

[Lasse Reichstein Nielsen, 2012/05/29 09:06:52]

Shouldn't there be a second argument here, probably str?
If this doesn't fail, is there a bug in assertEquals?

[Lasse Reichstein Nielsen, 2012/05/29 10:41:28]

My bad, you already fixed that.