🚀 go-pugleaf

steve@taumet.com (Steve Clamage) writes:
>darcy@druid.uucp (D'Arcy J.M. Cain) writes:
>>steve@taumet.com (Steve Clamage) writes:
>>>That said, suppose we have a system where int and size_t are 16 bits,
>>>but it is possible to allocate an object bigger than 64K.  A
>>But isn't size_t guaranteed to be big enough to specify any object?  It
>size_t is the type returned by the sizeof operator.  I believe you
>must be able to apply sizeof to any object you can declare.  Suppose

Mark Brader and I have been discussing this in email and I agree that
in the case of calloc (but not malloc) a stricly conforming compiler
might be able to allocate more bytes (but not elements) than can fit
into a size_t.  Here is an excerpt from my last message to him:

>Actually 4.10.3 never comes down on one side or the other but it does say
>that "Each allocation shall yield a pointer to an object disjoint from
>any other object."  Now of course you can't apply sizeof to the object
>because you only have a pointer to the object and hence you can only get
>the size of the pointer or the size of the first element.  However the
>discussion on sizeof does sort of imply that if you could somehow get
>an identifier for the object itself that sizeof could be applied to it
>and therefore size_t must be able to represent its size.  And of course
>if pigs had wings they might fly so I am not sure if I've proved anything
>other than that I come to these strange conclusions due to too much
>thought rather than too little.  :-)

-- 
D'Arcy J.M. Cain (darcy@druid.com)  |
D'Arcy Cain Consulting              |   There's no government
Toronto, Ontario, Canada            |   like no government!
+1 416 424 2871          DoD#0082   |