Fix y-umlaut to uppercase.

Fix y-umlaut to uppercase.

R=dcarney@chromium.org
BUG=v8:2984
Committed: https://code.google.com/p/v8/source/detail?r=17545

[Yang, 2013-11-06 15:55:36]

Please take a look.

I ended up not handlifying the code because there is no real need for it, and
because it would hurt microbenchmarks.

[dcarney, 2013-11-06 16:07:13]

lgtm

[Lasse Reichstein Nielsen, 2013-11-06 21:19:41]

drive-by comment :)

https://codereview.chromium.org/59853006/diff/50001/src/runtime.cc
File src/runtime.cc (right):

https://codereview.chromium.org/59853006/diff/50001/src/runtime.cc#newcode6209
src/runtime.cc:6209: // character is also ASCII.  This is currently the case,
but it
Is "ASCII" here still relevant?

[Sven Panne, 2013-11-07 07:07:24]

Another DBC...

https://codereview.chromium.org/59853006/diff/50001/src/runtime.cc
File src/runtime.cc (right):

https://codereview.chromium.org/59853006/diff/50001/src/runtime.cc#newcode6233
src/runtime.cc:6233: static const uc32 yuml_code = 0xff;
Hmmm, is this really the only case? Look e.g. at item 3 on
http://www.moserware.com/2008/02/does-your-code-pass-turkey-test.html, does this
apply here, too? I have to admit that I don't have any clue how (if at all) we
handle local-specific stuff here.

[dcarney, 2013-11-07 07:13:33]

https://codereview.chromium.org/59853006/diff/50001/src/runtime.cc
File src/runtime.cc (right):

https://codereview.chromium.org/59853006/diff/50001/src/runtime.cc#newcode6233
src/runtime.cc:6233: static const uc32 yuml_code = 0xff;
This code does not take locale into account.  It's only consistent with unibrow.
 The i18n extension does toUpperCase with locales

[Yang, 2013-11-07 08:49:36]

On 2013/11/06 21:19:41, Lasse Reichstein Nielsen wrote:
> drive-by comment :)
> 
> https://codereview.chromium.org/59853006/diff/50001/src/runtime.cc
> File src/runtime.cc (right):
> 
> https://codereview.chromium.org/59853006/diff/50001/src/runtime.cc#newcode6209
> src/runtime.cc:6209: // character is also ASCII.  This is currently the case,
> but it
> Is "ASCII" here still relevant?

In most cases, ASCII should be changed to OneByte after the change to Latin1,
and we still have some work here. But in this case, it is actually still
correct: this fast pass only handles the ASCII range, converting between a-z and
A-Z. Characters like Ä are not handled. In those cases, we simply drop to the
slow case afterwards.

[Yang, 2013-11-07 09:08:43]

Committed patchset #3 manually as r17545 (presubmit successful).