🚀 go-pugleaf

sthuang@hns.com (Steven Huang) in <4g1793$muh@solutions.solon.com>
wrote:

>You're way off track as well.  Short or long has nothing to do with
>the size of the integer in terms of bits.  You will never be guaranteed
>a certain number of bits you can use.  If I remember correctly the
>only guarantee you have is that short is less than or equal to int,
>and int is less than or equal to long.

The above is true only in the senses that (1) that the machine need not
have `bits' at all, and (2) there is no _maximum_ size for these types.
The standard requires that the implementation behave as if certain
things were true:

type        bits            min         max

char        >= 8            *           *       (same as either unsigned
signed "                    <=-127      >=127      or signed type)
unsigned "                              >=255

short                       <=-32767    >=32767
unsigned "                              >=65535

int (see short)

long                        <=-2147483647
                                        >= 2147483647
unsigned "                              >= 4294967295

It is very difficult to implement these limits on a binary machine
unless
char    >= 8 bits (mandated)
short   >= 16 bits
int     >= 16 bits
long    >= 32 bits

If, for example, sizeof(char) == sizeof(long), then a char would need to
be implemented as if it had 32 bits.

              
--
* Martin Ambuhl       net: mambuhl@ripco.com
* Chicago, IL (USA)