(No comment was entered for this change.)

I am not sure why you say it behaves like codePointAt -- JS still only has 16-bit
characters and uses surrogate pairs to represent non-BMP characters, just like Java.
 I have a patch to add support for the various codePoint routines in Character and
String, but they are definitely different functionality than charAt().

So, I am taking this issue as a request for adding codePoint* -- if you think there
is still some other difference then please post here.

Is there a compelling reason to support codePoint stuff for 1.5?

Is there a reason not to?  The code is pretty straightfoward and already written
(which took about an hour), and without it there is no way of supporting non-BMP
characters in GWT.  Asian languages in particularly make use of non-BMP characters,
so it seems we have a rather incomplete story if we leave it out.  Since part of the
1.5 push was to pick up 1.5 JRE additions, it seems like this needs to be there. 
Leaving it out only encourages developer to assume all characters are one char which
will break when their user types in a non-BMP character.

Granted, we are still missing classification tables for non-ASCII characters
(isLetter/etc, toLowerCase/etc), which is why I did not add the codePoint versions of
those.  I don't think that should go in 1.5 because it will take a lot more work to
craft the classification tables in such a way as to avoid bloating the generated JS.

Sure, that's totally fine, just didn't want this to take cycles from higher-priority
stuff.

Committed in r2185.

1_5_RC has been released.