Even in C it's not a literal. The consequence is that in a 32-bit implementation, you can't write INT_MIN as -2147483648 (-2^31), because 2147483648 would overflow a signed int. Instead a workaround such as -2147483647-1 must be used.
Not really. It's done in java. Usually, minus ("-") is parsed as a token and then unary minus on integer literals are transformed to a negative integer literal.
So, -2147483648 is read as MINUS INT_LITERAL and thus transformed into "-2147483648" as an INT_LITERAL, but -(2147483648) should be read as MINUS L_PAREN INT_LITERAL R_PAREN and should thus not be transformed.