Some inconsistencies observed with HarfBuzz output for Gujarati fonts

[behdad]

Discussed in #4281

^{Originally posted by sid-wr June 15, 2023}
Hi team,

I tried using hb-view tool for some Gujarati text & some broken clusters can be observed on checking the generated output.

The input text is -u "U+0AB8U+0ACDU+0AB5U+0ABEU+0AA4U+0A82U+0AA4U+0ACDU+0AB0U+0ACDU+0AAF"

The expected output is સ્વાતંત્ર્ય

I tried this for 20 different fonts & the Gujarati cluster is getting broken for the following 8 fonts:

1. Adobe Gujarati
2. Amdavadi Unicode
3. Bhavnagar Unicode
4. Bhuj Unicode
5. Rajkot Unicode
6. Surat Unicode
7. Vadodara Unicode
8. Vijaya Unicode

For the rest 12 fonts, the correct expected cluster is generated:

1. Skolar Gujarati
2. Gujarati Sangam
3. Kohinoor Gujarati
4. Nirmala UI
5. NotoSansGujaratiUI
6. NotoSerifGujarati
7. Ekatra
8. Anek Gujarati
9. Shruti
10. Padmaa
11. Rekha
12. Aakar

Also, I tried some other licensed third-party shaper software & this cluster is not getting broken for any of these 8 fonts at their end.
So can you please check whether the issue is at your end or at the respective fonts side ?

Have attached a combined screenshot of all outputs which I created for easy comparison:

Please help
Thanks

[behdad]

See discussion #4281 for my debugging results.

[behdad]

@jfkthame Should we hardcode the consonant positions for indie1?