Define a variable to distinguish system_icu from bundled_icu in Blink

Define a variable to distinguish system_icu from bundled_icu

Most of encoding name aliases manually added in TextCodecICU are present in Chrome's copy of ICU. So, they don't have to be added when Chrome's ICU is used.

Define 'USING_SYSTEM_ICU" when 'use_system_icu=1' and register the above encoding name aliases only when 'USING_SYSTEM_ICU' is set.

In addition, the following was done :

* Remove GBK aliases not specified in the encoding spec.
* Add tests for GBK and EUC-KR aliases in the spec that have
  not been tested before.
* Add two aliases for ISO-8859-8-I regardless of whether bundled or system
  ICU is used.
* Remove xA3xA0 => U+3000 override in GBK and GB18030 because
  Chrome's copy of ICU already has that with
    https://codereview.chromium.org/1162723008/
* Remove gbkFallback for 4 code points when bundled ICU is used for the same
  reason as above.
* Remove one-way mapping from Unicode for U+22EF and U+301C
  even with system_icu because Chinese IMEs on Mac do not produce them on Mac.  (they're added to Webkit to be compatible with an old Mac converter, but they're not specified in the encoding spec.)

BUG=493824
TEST=Layout test fast/encoding/*

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=196607

[jungshik at Google, 2015-06-01 23:49:49]

jshin@chromium.org changed reviewers:
+ jsbell@chromium.org

[jungshik at Google, 2015-06-01 23:49:49]

This CL works as intended with a bundled ICU, but I can't test a case with
'use_system_icu=1' because of http://crbug.com/341104. 

BTW, this has to be landed after https://codereview.chromium.org/1162723008/
(convrtrs.txt change; perhaps, I have to make a separate CL for that change
only).

[jsbell, 2015-06-02 00:06:32]

lgtm

https://codereview.chromium.org/1167523003/diff/40001/LayoutTests/fast/encodi...
File LayoutTests/fast/encoding/char-decoding.html (right):

https://codereview.chromium.org/1167523003/diff/40001/LayoutTests/fast/encodi...
LayoutTests/fast/encoding/char-decoding.html:13: //
<http://bugs.webkit.org/show_bug.cgi?id=17014> EUC-CN code A3A0 is mapped to
U+E5E5 instead of U+3000
Can we remove this comment? Looks like it passes now?

[jungshik at Google, 2015-06-02 00:20:35]

Thanks for the review.

https://codereview.chromium.org/1167523003/diff/40001/LayoutTests/fast/encodi...
File LayoutTests/fast/encoding/char-decoding.html (right):

https://codereview.chromium.org/1167523003/diff/40001/LayoutTests/fast/encodi...
LayoutTests/fast/encoding/char-decoding.html:13: //
<http://bugs.webkit.org/show_bug.cgi?id=17014> EUC-CN code A3A0 is mapped to
U+E5E5 instead of U+3000
On 2015/06/02 00:06:32, jsbell wrote:
> Can we remove this comment? Looks like it passes now?

Yes, it passes because the patch for the above bug manually added back the
mapping in the code. 

I think the comment is as to why this particular mapping is tested. 
Pending the resolution of GBK vs GB18030, I have not yet updated GBK/GB18030
tables to match the encoding spec more or less. When I do, I can remove the code
to map A3A0 to U+3000 (the patch for the above webkit bug).

[jungshik at Google, 2015-06-04 19:36:44]

Sorry for further changes after lgtm. 
Can you take another look?

Now I removed all the GBK-related overrides when ICU is bundled (i.e.
USING_SYSTEM_ICU is not defined).

[jungshik at Google, 2015-06-04 19:38:08]

jshin@chromium.org changed reviewers:
+ thakis@chromium.org

[jungshik at Google, 2015-06-04 19:38:09]

thakis@chromium.org: Please review/approve changes in Source/*. 
Thanks

[Nico, 2015-06-04 19:57:43]

do you need to change a BUILD.gn file too?

[jungshik at Google, 2015-06-04 23:05:32]

On 2015/06/04 19:57:43, Nico wrote:
> do you need to change a BUILD.gn file too?

gn does not yet support use_system_icu (according to a comment in
third_party/icu/BUILD.gn). I've just added a TODO comment to Source/BUILD.gn.

[Nico, 2015-06-04 23:09:02]

lgtm

[jungshik at Google, 2015-06-05 05:20:11]

Thanks, Nico. 

Joshua: are you ok with additional changes I made since your review? 

* Remove xA3xA0 => U+3000 override in GBK and GB18030 because
  Chrome's copy of ICU already has that with
    https://codereview.chromium.org/1162723008/
* Remove gbkFallback for 4 code points when bundled ICU is used for the same
  reason as above.
* Remove one-way mapping from Unicode for U+22EF and U+301C
  even with system_icu because Chinese IMEs on Mac do not produce them on Mac. 
(they're added to Webkit to be compatible with an old Mac converter, but they're
not specified in the encoding spec.)

[jsbell, 2015-06-05 16:34:30]

yes, still lgtm!

[jungshik at Google, 2015-06-05 20:08:37]

The CQ bit was checked by jshin@chromium.org

[jungshik at Google, 2015-06-05 20:08:37]

The patchset sent to the CQ was uploaded after l-g-t-m from thakis@chromium.org
Link to the patchset: https://codereview.chromium.org/1167523003/#ps160001
(title: "fix the comment in BUILD.gn")

[commit-bot: I haz the power, 2015-06-05 20:09:21]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1167523003/160001

[commit-bot: I haz the power, 2015-06-05 22:48:10]

Committed patchset #9 (id:160001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=196607