[HarfBuzz] TAI THAM SIGN SAKOT, Tone Marks and Syllable Boundaries

[Richard Wordingham]

Dear List,

I am raising this possible issue now because it is likely to be some
time before I have a font ready to check for a problem.

As those who have access to the Unicode members' list Unicore will
have read, in NFC/NFD Tai Tham text, a closed syllable with a tone mark
will usually end <SAKOT, tone, consonant>.  HarfBuzz 0.9.14
consequently identifies a cluster boundary between the tone and the
final consonant.  Does this intervening tone mark prevent the
conversion of the SAKOT and consonant to a subscript consonant?  I
believe some people here have Tai Tham fonts capable of testing this
aspect of HarfBuzz.

If there is a problem, I believe it can be solved by *internally*
assigning U+1A60 TAI THAM SIGN SAKOT a canonical combining class of 254.

If it's any consolation, my Graphite font also needs to be fixed to
cope with normalised text.

I've just been adding my own Unicode functions to update the system to
Unicode Version 6, and I've discovered some clustering anomalies:

(a) Before and after, _paimaa_ <LOW PA, AI, SAKOT, MA, AA> 'come and
go' is counted as a single cluster.

(b) Before and after, Unicode-normalised _cuewaa_ <LOW CA, UUE, SAKOT,
TONE-1, WA, AA> is counted as two clusters.

(b) Before I updated the Unicode functions, unnormalised _cuewaa_ <LOW
CA, UUE, TONE-1, WA, AA> was counted as one cluster.  After the update,
it is counted as two clusters.

Richard.