Red Hat Bugzilla – Bug 157815
tamil vowel signs "0BC1..0BC2 0BC6..0BC8 0BCA..0BCC" treated differently by icu and pango
Last modified: 2014-03-25 20:52:14 EDT
Description of problem:
Currently text selection, cursoring and text editing (using del key) have the
same problemdoes not highlight complete character in Indic. It is not consistent
with gtk+ application.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
2. enter 'kg'
3. move curosr back using '<-' key
3. move to the front of the character by pressing 'Home'
4. press delete once
3. shift '->' or '<-' to highlight the character
cursoring issue: takes two keypress to move across the character
delete issue: takes two keypress to delete the character
text selection: takes two keypress to hightlight the character
cursoring issue: one keypress as it is one composed character
delete issue: one keypress as it is one composed character
text selection: one keypress as it is one composed character
This is some sort of regression or change from 1.1.4 to 1.9.X in the icu module.
Affects upstream as well.
See http://www.openoffice.org/issues/show_bug.cgi?id=50172 for my testcase to
demo regression in icu 2.2 to icu 2.6
But also see http://www.jtcsv.com/cgibin/icu-bugs?findid=1587
According to the icu bug report this is the correct behaviour ? What do you
think, want me to try and revert the change anyway ?
The current http://www.unicode.org/reports/tr29/ says that The
Grapheme_Cluster_Break property values are defined in
that list does not list the tamil vowel signs, but the older
http://www.unicode.org/Public/3.2-Update/DerivedCoreProperties-3.2.0.txt did. So
*apparently* icu is following the spec.
I have a patch at http://qa.openoffice.org/issues/show_bug.cgi?id=50172 to
revert to the old OOo/icu behaviour which is the same as current pango and I'll
put it in into our OOO in the interim (i.e. 1.9.106-2)
any idea which of these behaviours is *correct* the pango one, or the icu one ?
I've matched the pango one for now.
Tested with openoffice.org-1.9.108-1, here's the breakdown.
bn,gu,hi,pa,ta: takes 1 keypress which is correct. all good.
bn,gu,hi,pa,ta: takes 1 keypress (Delete Key) which is correct. all good.
bn,gu,hi,pa,ta: takes 1 keypress (shift + ->) which is correct. all good.
Seems the width of character for bn, gu, hi are not right. Takes the space of
two characters instead of one. Please see attached screenshot, observe the
highlight of the character after i did shift+-> ONCE.
Created attachment 115282 [details]
screenshot with openoffice.org-1.9.108-1
character width issue.
character width issue should be the same problem as rh#157813#
putting the rendering issue aside as rh#157813# this works with 1.9.108-4 for