Bulletin number: 142 Products affected: D700D Description: Conversion of REAL constants by the compiler Component: THE COMPILER UTILITY SET Date: Tue May 15 17:00:00 GMT 1990 ----- The procedures used for converting a decimal real number into an IEEE floating point number representation within the compiler can get the least significant bit out by one in cases where it ought not (e.g (2**24 - 3)(REAL32)). This only arises for REAL numbers written as constants into \occam source, and not to constants converted at run-time by the library procedures STRINGTOREAL32 and STRINGTOREAL64 (and other procedures which call them) which perform the conversions to full accuracy.