[HarfBuzz] a problem in hangul normalization composed full

[Dohyun Kim]

Hi,

After having tested hangul shaping with malgun.ttf from Windows 8, I
realized that there is some problem in hangul composition process of
current hb-ng.

For instance, when input string is U+1109 U+1167 U+11F0, these
characters are composed into U+C154 U+11F0.  As a result, *jmo
features in malgun.ttf are not applied to this jamo sequence.

So it seems that we should not compose character sequence like the following:

modern hangul leading consonant (U+1100..U+1112)
+ modern hangul vowel (U+1161..U+1175)
+ pre-modern hangul trailing consonant (U+11C3..U+11FF, U+D7CB..U+D7FB)

When hb-ng meets these characters in a sequence, it should not compose
the first two characters into precomposed hangul syllable (U+AC00 ..
U+D7A3), but leave them as they are.

Regards,
-- 
Dohyun Kim
College of Law, Dongguk University
Seoul, Republic of Korea