[Freebangfont-devel] [Issue N22662] Bengali rendering bugs: followup

[Deepayan Sarkar]

Hi,

I just checked out the latest qt (rsync). Thanks for fixing the psts and ZWJ 
bugs. Unfortunately this has introduced a new bug related to handling of the 
combination ZWJ + ZWNJ (which used to work correctly before). This is 
important for the current unicode recommended encoding of khanda ta --- see 
attached files below for a example.


There's another bug in the 'init' lookup. The lookup is performed when the 
previous character satisfies !isLetter(). This holds for vowel signs, so the 
init form are substituted whenever preceded by a vowel mark. The following 
change seems to fix this:

deepayan $ diff qscriptengine_x11.cpp~ qscriptengine_x11.cpp
1536c1536
<       where[0] = (from == 0 || !string.unicode()[from-1].isLetter());
---
>       where[0] = (from == 0 || ! (string.unicode()[from-1].isLetter() || 
string.unicode()[from-1].isMark()));


Thanks again for your wonderful work.

Deepayan



P.S.: This is not terribly important, but I noticed the comment 


   // ### not sure if this is correct. If it is, does it apply only to Bengali 
or should
   // it work for all Indic languages?
   // the combination Independent_A + Vowel Sign AA is allowed.


I don't think (I may be wrong) this is any more correct than allowing other 
vowel signs after Independent_A.

bhartsiya-correct.png
Description: PNG image

bhartsiya-qtrsync.png
Description: PNG image

bhartsiya-utf8.txt
Description: Text document