ES6 regular expressions unicode mode: case insensitivity case folding inconsistency

[xqin]

• Version: v4.1.0
• Platform: Windows 7 x64

/(\.[\u4e00-\u9fa5A-Z])+/i.test('a.c');  //true

/(\.[A-Z\u4e00-\u9fa5])+/i.test('a.c');   //true

• Version: v6.3.0
• Platform: Windows 7 x64

/(\.[\u4e00-\u9fa5A-Z])+/i.test('a.c');  //false

/(\.[A-Z\u4e00-\u9fa5])+/i.test('a.c');   //true

why?

[vsemozhetbyt]

It seems here is the turning point:

/(\.[\u00ffA-Z])+/i.test('a.c'); // true
/(\.[\u0100A-Z])+/i.test('a.c'); // false

[mscdex]

I'm not familiar enough with the ES6 regexp changes, but FWIW adding the new /u flag makes the .test() return true.

[vsemozhetbyt]

Well, that is weird. Node.js 6.3.0

/(a[\u0100A-Z])+/i.test('aa'); \\ false
/a([\u0100A-Z])+/i.test('aa'); \\ true

/(a[\u0100A-Z])+/i.test('aa'); \\ false
/(a[\u0100A-Z])/i.test('aa');  \\ true

/(a[\u0100A-Z])+/i.test('aa'); \\ false
/(a[\u00ffA-Z])+/i.test('aa'); \\ true

/(a[\u0100A-Z])+/i.test('aa'); \\ false
/(a[A-Z\u0100])+/i.test('aa'); \\ true

/(a[\u0100A-Z])+/i.test('aa'); \\ false
/(a[\u0100a-z])+/i.test('aa'); \\ true

/(a[\u0100A-Z])+/i.test('aa'); \\ false
/(a[\u0100A-Z])+/i.test('AA'); \\ true

[targos]

Someone reported the bug upstream: https://bugs.chromium.org/p/v8/issues/detail?id=5199

[hashseed]

Issue has been fixed and ported back to V8 5.3 and 5.2. Please comment to the V8 issue if we need to merge this further back.

[ofrobots]

Thank! The older V8 streams (5.1 and older) are no longer accepting merges upstream, so fixes would have to be floated for master (V8 5.1) and v6.x (V8 5.0). I believe @fhinkel is planning to submit PRs for the floats.

[fhinkel]

Yes, working on it.