Inside GlyphPage::fill(), we create a string from a collection of codepoints in order to know if a particular font supports those codepoints. However, a string is fundamentally a different beast than an ordered collection of codepoints, because adjacent codepoints in a string are combined to form grapheme clusters. Therefore, this entire approach is incorrect. Luckily, we already have another approach. We just need to migrate to it in all cases.
*** This bug has been marked as a duplicate of bug 147920 ***