Pointer sizes

[lcc.dab at UCLA-LOCUS.ARPA]

From:            David Butterfield <lcc.dab at UCLA-LOCUS.ARPA>

First, I have seen it stated on this list that "char * is guaranteed by
the language to be the widest pointer".  I cannot seem to find this in
the K&R, however.  Reference to chapter and verse would be appreciated.

Second, K&R section 14.4, paragraph 4, sentence 3 says:

   "It is guaranteed that a pointer to an object of a given size may be
   converted to a pointer to an object of a smaller size and back again
   without change."

My question then is:  For purposes of the above sentence, what is the
size of a function relative to the data types.  I know how to find the
size of a data object using the sizeof operator.  I don't know how to
find the size of a function.

(Some might say that the "size" of a function is the number of "bytes"
between the closest possible starting points of functions.  On the
PDP-11, this would be 2, since functions must start on even addresses.)

The question becomes much worse, however, when functions and data live
in different address spaces with different sizes.  For example, consider
a Large Text, Small Data 8086 C compiler.  Here, a char * would
reasonably be 16 bits and a function * would have to be 32 bits.  Here,
functions would seem to be "smaller" than chars.

Worse yet would be a machine where different data types could live in
various different address spaces with different lengths, but lets not
get carried away.

(Some might say that for purposes of the above quote, the "size" of an
object is the fraction of its address space that it consumes.  In the
case of data, the fraction is computed using the length of the object.
In the case of functions, the fraction is computed using the closest
possible starting points.  This is yucky, but seems to reflect what size
of pointer is required.)