202883 - Line wrapping error in Korean text where a fullstop leads a line

Status: VERIFIED DUPLICATE of bug 164759

(Core :: Internationalization, defect)

Description

[Isaac Hwak Han]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.4a) Gecko/20030401
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.4a) Gecko/20030401

In the attached testcase page which has been created on purpose, all DIV boxes
have specified with of 560 pixel. The sentence of the first example is
incorrectly wrapped to two lines, where the second line begins with a fullstop
('.').

Reproducible: Always

Steps to Reproduce:
1. open the testcase page in Mozilla
2. See the result

Actual Results:  
The second line of the first example begins with a fullstop.

Expected Results:  
The sentence should be wrapped just before the character (다, U+B2E4) preceding
the fullstop.

When the fullstop is replaced with an exclamation mark or other punctuation
mark, the browser shows the correct wrapping.

This does not occur in English text, only but in Korean text.

Comment 1

[Isaac Hwak Han]

Attachment: Testcase

Comment 2

[Jo Hermans]

*** Bug 202884 has been marked as a duplicate of this bug. ***

Comment 3

[Isaac Hwak Han]

This may be a Korean variation of Bug 164759.

Comment 4

[Yuying Long]

Confirm with 04-21 trunk build on WinXP.

Comment 5

[Yuying Long]

Attachment: Test case 2

Another test case:
Delete/add the single byte space in 1st and 3rd string in original test case
has different result.

Comment 6

[Yuying Long]

Please ignore the 3rd case in previous test case 2 - I copy/pasted the problem
1st string into 3rd string, I can see the problem in Composer, but when bring it
in Browser window, the line wrap shows fine. 

Comment 7

[Jungshik Shin]

Attachment: a bit more extenstive test case

I think this and bug 164759 are identical 

For CJK linebreaking (break at any grapheme boundaries except where prohibited 

per JIS X 4501), JIS X 4501 is used, I believe. I put comma and right
parenthesis 
in place of full stop and the result is the same for all three characters that
belong to JIS X 4501 class 'a' (actually, in JIS X 4501, full stop
is class 6 and comma and right parenthesis are class 2, but both
classes behave indetically so that in Mozilla they're combined into
class 'a').  (see
http://lxr.mozilla.org/seamonkey/source/intl/lwbrk/src/jisx4501class.h
and related files in the directory)

Comment 8

[Jungshik Shin]

Exclamation mark(U+0021) is  class 'a' as are U+0029, U+002C and U+002E 
('1' in the following bit mask)
 
0x88438815, // U+0020 - U+0027
0x81515810, // U+0028 - U+002F

I have to look more closely why '!' (for that matter '?') behaves differently 
from others in class 'a'. It may not be the line breaker but its caller
(e.g.
http://lxr.mozilla.org/seamonkey/source/layout/html/base/src/nsTextTransformer.cpp#550)
that needs to be looked at.

Comment 9

[Jungshik Shin]

I don't know what's into my head when I wrote that comma and right parenthesis
behave the same way as full stop in comment #7. They don't. (so ignore my
comment #8) Anyway, I found the cause and have a patch. Let's move this to bug
164759
because there's nothing Korean specific here. 



*** This bug has been marked as a duplicate of 164759 ***

Comment 10

[Yuying Long]

Mark as verified as dup, it's sure not Korean specific.