JTC1/SC22/WG14 N629

                          Document Number:  WG14 N629/X3J11 96-093
       
       
                               C9X Revision Proposal
                               =====================
       
       Title: ____Type rules for decimal integer constants_________
       Author: ___David Prosser____________________________________
       Author Affiliation: __SCO, Inc._____________________________
       Postal Address: 430 Mountain Ave.; P.O. Box 4; Murray Hill, NJ 07974
       E-mail Address: [email protected]___________________________
       Telephone Number: ____(908)790-2358_________________________
       Fax Number: __________(908)790-2426_________________________
       Sponsor: __Larry Jones______________________________________
       Date: ________________13 November 1996______________________
       Proposal Category:
          __ Editorial change/non-normative contribution
          x_ Correction
          x_ New feature
          __ Addition to obsolescent feature list
          __ Addition to Future Directions
          __ Other (please specify)  ______________________________
       Area of Standard Affected:
          __ Environment
          x_ Language
          __ Preprocessor
          __ Library
             __ Macro/typedef/tag name
             __ Function
             __ Header
          __ Other (please specify)  ______________________________
       Prior Art: _________________________________________________
       Target Audience: _C Programmers_____________________________
       ____________________________________________________________
       ____________________________________________________________
       Related Documents (if any): ________________________________
       ____________________________________________________________
       ____________________________________________________________
       Proposal Attached: x_ Yes __ No, but what's your interest?

       Abstract:
	The addition of a new integer data type (long long) permits a
	change to the rules regarding types for decimal integer constants
	that fixes some inconsistencies, primarily regarding signedness.

	This proposal strongly suggests that decimal integer constants
	should only have signed integer type choices, unless suffixed by
	a "u" or "U".

       Proposal:

	The rules for decimal integer constants in the current ISO C
	standard reflect the notion that a long is the biggest type and,
	having predated the "u" and "U" suffix addition, had decimal
	integer constants whose value was in the range [LONG_MAX+1,ULONG_MAX]
	taken to have type unsigned long.  Otherwise, an unsuffixed decimal
	constant must be a signed type.  For the other bases (octal and
	hexadecimal), all the signed and unsigned integer types from int
	and up are sequentially stepped through.

	When long long was added, integer constants had to be accepted
	with values beyond that representable by long or unsigned long.
	With octal and hexadecimal, adding long long and then unsigned
	long long is straightforward and obvious.  The technical committee
	now had to wrestle with the problem of these decimal constants.
	The two choices where either to similarly tack on long long and
	then unsigned long long, or first to remove the unsigned long
	choice and then tack on the same two new types.  My understanding
	is that there is existing practice in both of these camps.  I
	know that SCO's C compiler took the latter choice (and emits a
	warning when a decimal integer constant without a "u" or "U"
	suffix falls in the above range).

	What came out of a discussion partially seen on the sc22wg14
	reflector is a different choice--one that we probably would have
	strongly considered back in the 80's when we were first codifying
	C (even though there we had no long long type) if we had seen the
	expressiveness of our "u" and "U" suffix addition.  This choice
	is simply not to include ANY unsigned integer types for decimal
	integer constants without a "u" or "U" suffix.  In other words,
	a constraint violation (ISO C 6.1.3) would be triggered if such
	a constant had a value greater than LLONG_MAX.

	[N.B.: I've assumed that <limits.h> has LLONG_MIN, LLONG_MAX and
	ULLONG_MAX as its spelling of the new end point values.  These are
	the spellings used in SCO's implementation.]

	The reworded ISO C 6.1.3.2 "semantics" second paragraph would read:

	The type of an integer constant is the first of the corresponding
	list in which its value can be represented.  Unsuffixed decimal:
	int, long int, long long int; unsuffixed octal or hexadecimal: int,
	unsigned int, long int, unsigned long int, long long int, unsigned
	long long int; suffixed by the letter "u" or "U": unsigned int,
	unsigned long int, unsigned long long int; decimal suffixed by the
	letter "l" or "L": long int, long long int; octal or hexadecimal
	suffixed by the letter "l" or "L": long int, unsigned long int,
	long long int, unsigned long long int; suffixed by both the letters
	"u" or "U" and "l" or "L": unsigned long int, unsigned long long int;
	decimal suffixed by two letters "l" or "L": long long int; octal or
	hexadecimal suffixed by two letters "l" or "L": long long int,
	unsigned long long int; suffixed by both "u" or "U" and two letters
	"l" or "L": unsigned long long int.

	(I would suggest that this can be much more easily presented now
	that long long exists by a table!)