Combining characters break Char literals

Kotlin's Char is documented, precisely, as "a 16-bit Unicode character" — one UTF-16 code unit. A single letter a reader perceives as one character can need two of them: a base letter followed by a separate combining mark that is rendered on top of it. क़ (Devanagari "qa") is क (KA) plus a combining nukta sign — two code points, two Chars — and 'क़' does not compile as a character literal at all. The same wall applies, for an unrelated reason, to any single code point outside the Basic Multilingual Plane: it needs a surrogate pair to represent in UTF-16, so it is also two Chars masquerading as one glyph — see kmp-html-entity-decoder for the worked branch that turns a decoded code point above 0xFFFF back into that pair.

Traps

A map keyed by Char cannot hold an entry for a combining sequence — the literal itself fails to compile. Perso-Arabic loanword sounds in Hindi are written as a base consonant plus a nukta; the source here handles it not by trying 'क़' to "q" (which is rejected outright) but by keeping the nukta as its own Char constant and treating it as a modifier the lookup consumes separately: