Noto Sans Kannada document fonts misrendering RA+VIRAMA+RA

[GoogleCodeExporter]

I found one immediate minor bug. The sequence 0CB0 + 0CCD + 0CB0 is rendered 
wrongly. I have attached the screenshot depicting the wrong and correct display 
forms. 

Thanks and regards,
Pavanaja

Original issue reported on code.google.com by pavan...@gmail.com on 17 Jul 2013 at 11:38

Attachments:

• NotoSansKannada-Wrong-display-Jul-16-2013.jpg

[GoogleCodeExporter]

Thank you for the report. What program did you use to render the above and on 
what platform? 

Original comment by jungs...@google.com on 15 Aug 2013 at 12:25

[GoogleCodeExporter]

Original comment by tha...@google.com on 15 Aug 2013 at 12:30

[GoogleCodeExporter]

I'm seeing the same result that pavanaja is calling "wrong display" here with 
the latest versions of the Kannada font and latest version of HarfBuzz. So this 
is not rendering engine related.

The character sequence is RA + VIRAMA + RA.

Cibu, can you take a look?

Original comment by roozbeh@google.com on 20 Sep 2013 at 9:46

[GoogleCodeExporter]

Confirming that the behavior of Window's Tunga font is different from Noto's, 
and it produces the rendering described as "correct display". It still may be a 
HarfBuzz bug, but less probably so.

Original comment by roozbeh@google.com on 20 Sep 2013 at 9:58

[GoogleCodeExporter]

Original comment by roozbeh@google.com on 20 Sep 2013 at 9:58

• Changed title: Noto Sans Kannada potentially misrendering RA+VIRAMA+RA

[GoogleCodeExporter]

I don't think this is useful sequence; meaning, no words exists with this 
sequence. So display either way should be fine. Please reopen this issue with 
some evidence, if my guess is incorrect.

Original comment by c...@google.com on 21 Sep 2013 at 1:59

• Changed state: Invalid

[GoogleCodeExporter]

It seems to me that a Kannada transcription of "cherry" uses the sequence: 
ಚೆರ್ರಿ (CA + E + RA + VIRAMA + RA + I). And again, HarfBuzz renders 
the string differently with Noto Sans and Tunga. Attaching renderings for both.

Some randoms searches also point me to some transcriptions for "Terra" and 
"Andorra".

Back to Cibu.

Original comment by roozbeh@google.com on 21 Sep 2013 at 2:18

• Changed state: Accepted

Attachments:

• cherry-noto.png
• cherry-tunga.png

[GoogleCodeExporter]

There are many more words which use this sequence. For example "sarrane", ie., 
ಸರ್ರನೆ

Original comment by pavan...@gmail.com on 21 Sep 2013 at 3:10

[GoogleCodeExporter]

Confirming that HarfBuzz renders ಸರ್ರನೆ differently with Noto and 
Tunga too.

Original comment by roozbeh@google.com on 21 Sep 2013 at 5:56

[GoogleCodeExporter]

I agree with Pavanaja and Roozbeh that, this is an issue and needs to get 
passed on to the designer to allow this exception for reph handling.

Original comment by c...@google.com on 23 Sep 2013 at 7:25

[GoogleCodeExporter]

I checked GSUB table in Tunga font and it has got ರ್ರ specifically taken 
care of. Probably something like has to happen in Noto.

Original comment by c...@google.com on 30 Sep 2013 at 8:42

[GoogleCodeExporter]

Cibu, would you please open a noto-alpha bug for this?

Original comment by roozbeh@google.com on 26 Dec 2013 at 9:20

[GoogleCodeExporter]

Any update on this?

-Pavanaja

Original comment by pavan...@gmail.com on 19 Mar 2014 at 5:26

[GoogleCodeExporter]

Sorry, I had forgotten to inform the designer about this bug. Just did that. 
Thanks for reminding..

Original comment by c...@google.com on 19 Mar 2014 at 5:39

[GoogleCodeExporter]

The designer has extended the render rules in the font to deal with this 
problem. The fix will make it into the next font update. 
When the shaper software encounters the ambiguious combination RA-virama-RA it 
has to decide whether to apply the below base form feature to the second RA or 
to re-order the first and apply the reph feature. It turns out it is required 
for each font to contain an eleborate contextual substitution and position swap 
feature to correct the choice of the shaper.

Original comment by jelle.bo...@monotypeimaging.com on 19 Mar 2014 at 10:05

[GoogleCodeExporter]

For the record, the internal tracking issue is noto-alpha/187.

Original comment by roozbeh@google.com on 8 May 2014 at 4:25

[GoogleCodeExporter]

This seems to be fixed in the latest UI fonts but not the document fonts.

Original comment by roozbeh@google.com on 18 Jul 2014 at 5:33

• Changed title: Noto Sans Kannada document fonts misrendering RA+VIRAMA+RA

[GoogleCodeExporter]

This was fixed in August with 
https://code.google.com/p/noto/source/detail?r=7075ada27edbf6eff9aa167a22cf05bf7
8552287

Original comment by roozbeh@google.com on 1 Nov 2014 at 11:13

• Changed state: Fixed

[GoogleCodeExporter]

The sequence 0CB0 + 0CCD + 0CB0 is still rendered wrongly. I just now checked. 
I am using Windows 8.1. Checked in Notepad and MS Word 2010. 

Original comment by pavan...@gmail.com on 23 Mar 2015 at 6:38

[GoogleCodeExporter]

Are you using the latest fonts? I'm using version 1.03. I don't have Windows, 
but with HarfBuzz, <0CB0, 0CCD, 0CB0> renders the same way as Tunga.

Original comment by roozbeh@google.com on 23 Mar 2015 at 4:33

[GoogleCodeExporter]

The font that I tested was downloaded today from 
http://www.google.com/get/noto/#/. It is version 1.02. Where is ver 1.03?

Original comment by pavan...@gmail.com on 23 Mar 2015 at 4:48

[GoogleCodeExporter]

At 
https://code.google.com/p/noto/source/browse/#git%2Ffonts%2Findividual%2Fhinted 
and 
https://code.google.com/p/noto/source/browse/#git%2Ffonts%2Findividual%2Funhinte
d

Original comment by roozbeh@google.com on 23 Mar 2015 at 6:01

[GoogleCodeExporter]

It is downloading huge file, probably includes all fonts. How do I download 
only Kannada fonts?

Original comment by pavan...@gmail.com on 24 Mar 2015 at 2:41

[GoogleCodeExporter]

Hi Pavanaja, here is the direct link: 
https://noto.googlecode.com/git/fonts/individual/unhinted/NotoSansKannada-Regula
r.ttf

Original comment by c...@google.com on 25 Mar 2015 at 3:56