Bhaiksuki and Grantha USE overrides

[punchcutter]

For Bhaiksuki the two gap fillers U+11C44 and U+11C45 should be treated like letters so they can allow a following visarga. Here's manuscript and Noto as reference:

For Grantha the anusvara U+11302 currently has a USE category VMPst, but it needs to allow Vedic marks like udatta U+0951 and anudatta U+0952 to follow. Those are VMAbv and VMBlw. It also needs to allow the Sama Vedic svaras which are all VMAbv so I suggest setting the anusvara to VMAbv as well.

[ebraminio]

Let me see if I can improve this :)

Can you put the exact text in Unicode string? That would be nice to put in test suit eventually I guess.

[punchcutter]

For Bhaiksuki the string above is: 𑰧𑱅𑱅𑰾𑱁 11C27,11C45,11C45,11C3E,11C41
You can use Noto Bhaiksuki to see that work.
Grantha with anusvara and Vedic mark:𑌷𑌂𑍧 11337,11302,11367 or 𑌕𑍍𑌷𑌂॒ 11315,1134D,11337,11302,0952
Noto Grantha isn't out yet to test that, but I verified setting U+11302 to VMAbv works. I updated hb-ot-shape-complex-use-table.cc and tested. I was going to just send a pull request updating gen-use-table.py, but as Behdad says in #539 these overrides should really be put into data files.

[ebraminio]

Ok. I see that changing 11C3E from VMPst to either of {H,B,O} fixes the Bhaiksuki case. So should we just wait for the next unicode update or is there anything we need to do here?

[dscorbett]

The fix for Bhaiksuki should not be to change U+11C3E (the visarga) but to change the two gap fillers to Indic_Syllabic_Category=Consonant_Placeholder.

[punchcutter]

Yes, U+11C3E is not the problem. I submitted a pull request. Some tests configs seem to be failing, but that seems unrelated.

[ebraminio]

Ok so. Thanks!