Implemented invalid UTF8 detection in decodeURI.

Implemented invalid UTF8 detection in decodeURI.  That is, detection
of invalid utf8 not invalid utf8-detection.

[Christian Plesner Hansen, 2009-03-09 17:58:53]

Oh and I forgot to mention: fixed a tiny inconsistency wrt. regexp function
lengths.

[Lasse Reichstein, 2009-03-10 07:25:03]

LGTM.

http://codereview.chromium.org/40300/diff/1/2
File src/regexp-delay.js (right):

http://codereview.chromium.org/40300/diff/1/2#newcode304
Line 304: 
Was the comment wrong, or has Firefox changed to 1?

http://codereview.chromium.org/40300/diff/1/3
File src/uri.js (right):

http://codereview.chromium.org/40300/diff/1/3#newcode93
Line 93: var value;
How about assigning octets[0], octets[1] and octets[2] to variables, instead of
doing so many array lookups?
Is this auto-generated code?

http://codereview.chromium.org/40300/diff/1/4
File test/mjsunit/regress/regress-244.js (right):

http://codereview.chromium.org/40300/diff/1/4#newcode61
Line 61: assertTrue(e instanceof URIError);
Please add a message to assertTrue calls to identify the erroring test. In this
case, perhaps just the value 
 var message = value + ": threw " + e;

http://codereview.chromium.org/40300/diff/1/4#newcode64
Line 64: assertTrue(threw);
Another way to make this assertion, without the need for a variable, is to
insert a 'fail(value);' call at the end of the body of the try-statement.

[Christian Plesner Hansen, 2009-03-10 07:46:52]

http://codereview.chromium.org/40300/diff/1/2
File src/regexp-delay.js (right):

http://codereview.chromium.org/40300/diff/1/2#newcode304
Line 304: 
The comment was wrong in the sense that the spec does say something about the
length of exec and test.  I don't anticipate any compatibility issues.

http://codereview.chromium.org/40300/diff/1/3
File src/uri.js (right):

http://codereview.chromium.org/40300/diff/1/3#newcode93
Line 93: var value;
You only know if octets[2] exists when you've inspected octets[0] and octets[1]
so you can't declare those variables up front and declaring them along the way
makes the code more difficult to read.  But I'll see if I can refactor it
somehow to read properties less often.

It was hand written, thank you very much, but with no attempts of being smart
about anything because it's so easy to get wrong.

http://codereview.chromium.org/40300/diff/1/4
File test/mjsunit/regress/regress-244.js (right):

http://codereview.chromium.org/40300/diff/1/4#newcode61
Line 61: assertTrue(e instanceof URIError);
Fixed

http://codereview.chromium.org/40300/diff/1/4#newcode64
Line 64: assertTrue(threw);
Fixed