391045 - firefox fail to render U+0D28 U+0D46 (Malayalam NA E) properly

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Biju]

Attachment: thalli.html

occasionally firefox fail to render malayalam നെ properly in word തള്ളിയതിനെ
see attachments thalli.html and thalli.png

Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9a8pre) Gecko/2007080505 Minefield/3.0a8pre

Comment 1

[Biju]

Attachment: thalli.png

Comment 2

[Biju]

Attachment: thalli.html

fix issue with previous upload of thalli.html

Comment 3

[Simon Montagu :smontagu]

Attachment: Reduced test case

Any attempt to minimize the testcase further than this makes the bug go away.

Comment 4

[Simon Montagu :smontagu]

We are calling ScriptShape() for all the ranges in the item with the same SCRIPT_ANALYSIS structure. This means that the fLinkBefore and fLinkAfter flags, which only refer to the beginning and end of the item respectively, are being applied to each range. This can potentially cause misrendering in any complex text string and should be a blocker.

Comment 5

[Simon Montagu :smontagu]

Attachment: Patch

Comment 6

[Simon Montagu :smontagu]

Attachment: Rendering with the patch

Biju, can you take a look at this screen shot of the testcase with my patch applied? I see that the word തുടര്‍ന്നാണ് is also rendered differently (but the same as in Firefox 2 and IE) and I'd like to be sure that it's correct, and that there are no other errors introduced.

Comment 7

[Biju]

(In reply to comment #6)
> Biju, can you take a look at this screen shot of the testcase with my patch
> applied? I see that the word തുടര്‍ന്നാണ് is also

I assume you are using either Karthika(Kartika) or AnjaliOldLipi as your Malayalam font,
then rendering of തുടര്‍ന്നാണ് in attachment 275756 [details] is CORRECT

Details:-

<QUOTE From = http://unicode.org/review/pr-66.html >

In The Unicode Standard, Version 4.0, chillu forms and explicit virama forms are distinguished in text by the use of joiners. For example:

    <NA, VIRAMA, ZWNJ>  to represent a NA with a visible virama

    <NA, VIRAMA, ZWJ>  to represent the chillu N

However, the standard is not clear about what <NA, VIRAMA> (without any joiner) represents. Some readers have interpreted the standard such that this sequence represents NA with a visible virama, others have interpreted it to be the chillu N, and yet another possibility is that the sequence does not specify one or the other form (i.e. the rendering system can choose).

</QUOTE>

So Karthika and AnjaliOldLipi render <NA, VIRAMA> is rendered as  <NA, VIRAMA, ZWNJ>  

And in ThoolikaTraditionalUnicode from SuperSoft
http://www.supersoftweb.com/Unicode.htm gives some prob with firefox and
(i think without some proper windows ddl)
and <NA, VIRAMA> is rendered as  <NA, VIRAMA, ZWJ>  

Confusion with <const+VIRAMA> also exist for other "chill" letters
I am listing all "chill"s here for quick reference
<const>, <const+VIRAMA>, <const+VIRAMA+ZWNJ>, <const+VIRAMA+ZWJ>

ന ന് ന്‌ ന്‍ 
ണ ണ് ണ്‌ ണ്‍
ര ര് ര്‌ ര്‍
ല ല് ല്‌ ല്‍
ള ള് ള്‌ ള്‍

future chill
ക ക് ക്‌ ക്‍
യ യ് യ്‌ യ്‍


(In reply to comment #6)
> and that there are no other errors introduced.

I see no other problem in attachment 275756 [details] 

Comment 8

[Simon Montagu :smontagu]

(In reply to comment #7)

> I assume you are using either Karthika(Kartika) or AnjaliOldLipi as your
> Malayalam font,
> then rendering of തുടര്‍ന്നാണ് in attachment 275756 [details] is
> CORRECT

Yes, I'm using AnjaliOldLipi. Thanks for testing!

Comment 9

[Biju]

When will this patch make into the nightly build?

PS: You can understand how URGENT this is by seeing edit/reedit of േകരളം <=> കേരളം at
http://en.wikipedia.org/w/index.php?title=Kerala&diff=151994033&oldid=151980972

Since years IE render it out of box correctly, 
so I can guarantee that all those users making it incorrect are Firefox (version < 3) users
And that is a kind of shame for us.

see also http://en.wikipedia.org/wiki/Help:Multilingual_support_%28Indic%29

Comment 10

[Simon Montagu :smontagu]

Checked in.

Comment 11

[Biju]

(In reply to comment #10)
> Checked in.

tested....
so far looks good, thanks Simon ....