padpowell@wateng.UUCP (PAD Powell) (10/24/84)
I have just been burned again, by a C compiler, and a structure.
I propose the following construct to be added to C:
offsetof( structure_name.field.field.... )
Given the name of the stucture, and the list of the fields, returns the
integer value of the offset from the beginning of the structure, that
will be used in accessing the particular field.
1. This is useful in DEBUGGING, and finding if you have portable code.
I have just been burned with 16/32 bit integers.
2. It provides a "hook" into the compiler which will eliminate all the
ugly arguements about "offset calculations using 0 pointers"
which is raging.
3. In my personal opinion, it is a solution to a problem that requires
looking at object code, and then guessing what the offset is.
Patrick ("Whimper! first the optimizer gets me, now the structures")
Powellmarcus@pyuxt.UUCP (M. G. Hand) (10/26/84)
Well, if you really want to know what the actual offset into your data structure is (as opposed to some logical naming method like struct.mem) you should try using BCPL instead of C. Try this: LAST.LIST!NEXT := NEWVEC(SIZE.OF.LIST) LIST.MEM := LAST.LIST!NEXT PACKSTRING(LIST.MEM!NAME, "a string") where NEXT and NAME are simple integers giving offsets into the vector returned by NEWVEC. Ie you get a simple contiguous block of BCPL words and its is up to you how you partition it. To get machine portability you are guaranteed the manifets constants BCPLBYTESPERWORD and BITSPERBYTE which can be used as scaling factors. Fun, eh? marcus