REGRESSION(130231): Leaves m_glyphToCharacterIndexes uninitialized

[Dominik Röttsches (drott)]

As commented in https://bugs.webkit.org/show_bug.cgi?id=97993#c13
the patch from r130231 removes

-    hb_glyph_info_t* infos = hb_buffer_get_glyph_infos(harfBuzzBuffer, 0);
-    for (unsigned i = 0; i < m_numGlyphs; ++i)
-        m_glyphToCharacterIndexes[i] = infos[i].cluster;

in HarfBuzzShaper::HarfBuzzRun::applyShapeResult.

And I can't see any replacement for that in r130231. AFAICS, with that, m_glyphToCharacterIndexes becomes uninitialized.

Unfortunately, I don't have a test to prove it yet. But I am suspecting it breaks selection in certain cases.

[Dominik Röttsches (drott)]

Created attachment 196364 [details]
Patch

[Dominik Röttsches (drott)]

Never mind, it is initialized in

@@ -361,6 +346,7 @@ void HarfBuzzShaper::setGlyphPositionsForHarfBuzzRun(HarfBuzzRun* currentRun, hb
     hb_glyph_position_t* glyphPositions = hb_buffer_get_glyph_positions(harfbuzzBuffer, 0);
 
     unsigned numGlyphs = currentRun->numGlyphs();
+    uint16_t* glyphToCharacterIndexes = currentRun->glyphToCharacterIndexes();

[Brent Fulgham]

Comment on attachment 196364 [details]
Patch

Clearing review patch since this was resolved.