The byte sequences "0x27 0x0a 0x27", "0x27 0x0d 0x27", and "0x27 0x27 0x27" (newline, carriage return, and single-quote, respectively, sandwiched between single quotes) are accepted as character literals. The former two are, as far as I can tell, allowed per the manual's description of the language, but would not feature in any sane language; I assume this is merely an oversight. The latter is rejected by the grammar described in the manual but accepted by the compiler. Presumably the manual is the authority and the compiler is wrong to accept '''.