Re: another question about layout in ICU 2.6

Hi Jens,

I've looked at all the related fixes I can think of that I've made to 
the code since 2.6, and I found some that seem like they'd be relevant, 
but none of that code seems to be hit when I run the latest code using 
your sample text and the Khmer OS font.

Most of the changes related to broken fonts. There was one on line 37 of 
DeviceTables.cpp that I was sure would be it - change:

             result |= ! fieldMasks[format];

to:

             result |= ~ fieldMasks[format];

(i.e. replace "!" with "~")

I put a breakpoint on this line in the most recent ICU, and ran your 
sample with the Khmer OS font, and this line was never hit... You might 
try this patch anyhow just to see if it helps - it's easy enough to 
make. :-)

Other than that, I have to say that I'm stumped. You might try looking 
at all the positioning-related differences between your patched 2.6 code 
and the released 3.2 code to see if anything jumps out at you.

(It might be a better use of everyone's time to just move to ICU 3.2 - 
that will get you a lot of other improvements as well ;-)

Regards,
Eric


Jens Herden wrote:
> Hi Eric,
> 
> 
>>I've been looking at the changes I've made to the LayoutEngine since ICU
>>2.6, and I'm not sure that any of them address the problem you describe.
> 
> 
> I very much appreciate your support here, thanks.
> 
> I want to say again, all this is not an issue with ICU 3.2. It is only visible 
> with 2.6 and I took the version without the patch from OOo to be sure it is 
> really causes by ICU. 
> 
> 
>>Can you please send me an example of a character sequence that is
>>rendering incorrectly? Also, let me know if you see the problem with all
>>Khmer fonts, or only certain fonts. I'll see if I can figure out what
>>other patches you might need to make it work.
> 
> 
> You may download:
> 
> ftp://ftp.khmeros.info/download/issues/Sample.txt
> ftp://ftp.khmeros.info/download/issues/correct-kampongtrach.ttf.png
> ftp://ftp.khmeros.info/download/issues/wrong-khmeros.ttf.png
> 
> to get an idea what my problem is. 
> 
>>From inspecting the fonts I can tell you that there is a major difference how 
> they define the subscript glyphs. In kampongtrach.ttf happens a simple 
> substitution from coeng + consonant -> subscript consonant. In khmeros.ttf 
> there are also anchors defined to finetune the position. 
> My guess is the calculation with the anchors is broken. 
> 
> 
>>One major change I did since 2.6 was to implement the Multiple
>>Substitution subtable - it's possible that this is somehow related,
>>although this code is not hit in ICU 3.2 using the Khmer samples and
>>fonts that I have.
> 
> 
> I can not image how this would cause what I see.
> 
> Thanks again
> 
> Jens
> _______________________________________________
> icu mailing list
> [email protected]
> http://oss.software.ibm.com/developerworks/oss/mailman/listinfo/icu
> 
> 
>