Detect clusters in Indic scripts where an initial RA is reordered after the following consonant and vowel. Bug 428317, r=roc

This commit is contained in:
Simon Montagu 2008-11-06 02:02:51 -08:00
Родитель 0188432e2c
Коммит 88115cd125
1 изменённых файлов: 16 добавлений и 6 удалений

Просмотреть файл

@ -1227,16 +1227,26 @@ PostLayoutCallback(ATSULineRef aLine, gfxTextRun *aRun,
while (glyphCount < numGlyphs) {
ATSLayoutRecord *glyph = &glyphRecords[glyphIndex + direction*glyphCount];
PRUint32 glyphOffset = glyph->originalOffset;
PRUint32 nextIndex = isRTL ? glyphIndex - 1 : glyphIndex + 1;
PRUint32 nextOffset;
if (nextIndex >= 0 && nextIndex < numGlyphs) {
ATSLayoutRecord *nextGlyph = &glyphRecords[nextIndex + direction*glyphCount];
nextOffset = nextGlyph->originalOffset;
}
else
nextOffset = glyphOffset;
allFlags |= glyph->flags;
if (glyphOffset <= lastOffset) {
if (glyphOffset <= lastOffset || nextOffset <= lastOffset) {
// Always add the current glyph to the ligature group if it's for the same
// character as a character whose glyph is already in the group,
// or an earlier character. The latter can happen because ATSUI
// sometimes visually reorders glyphs; e.g. DEVANAGARI VOWEL I
// can have its glyph displayed before the glyph for the consonant that's
// it's logically after (even though this is all left-to-right text).
// In this case we need to make sure the glyph for the consonant
// is added to the group containing the vowel.
// sometimes visually reorders glyphs. One case of this is that DEVANAGARI
// VOWEL I can have its glyph displayed before the glyph for the consonant
// that it's logically after (even though this is all left-to-right text).
// Another case is that a sequence of RA; VIRAMA; <consonant> ; <vowel> is
// reordered to <consonant> ; <vowel> ; RA; VIRAMA.
// In these cases we need to make sure that the whole sequence of glyphs is
// processed as a single cluster.
} else {
// We could be at the end of a character group
if (glyph->glyphID != ATSUI_SPECIAL_GLYPH_ID) {