I'm reading the specification on bitwise shift operators, and I'm unsure whether integers are ever required to be two's complements.
In NumberBitwiseOp, it's explicitly mentioned that
Let lbits be the 32-bit two's complement bit string representing ℝ(lnum).
Which probably means that for &, |, and ^, I can confidently say that the operation is working on 32-bit two's complements.
In BigInt::leftShift, it's mentioned that
Semantics here should be equivalent to a bitwise shift, treating the BigInt as an infinite length string of binary two's complement digits.
Which implies that bitwise shift should operate on two's complements.
However, in Number::leftShift, it's only mentioned that
Return the result of left shifting lnum by shiftCount bits.
Where lnum is the result of ToInt32, the latter of which is defined by the mathematical value, not by its binary encoding. Unlike NumberBitwiseOp, the operation here is working on a number, not a bit string, and the encoding is not specified. The next sentence says " The mathematical value of the result is exactly representable as a 32-bit two's complement bit string", but that sounds like an invariant assertion instead of a normative specification.
Is that an editorial mistake, or have I overlooked something? Does "left shift" necessarily imply "two's complement"?