From -6089938785496519169 X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: f78e5,e94695845965b249 X-Google-Attributes: gidf78e5,public X-Google-ArrivalTime: 1992-10-23 09:11:50 PST Newsgroups: comp.std.c++ Path: sparky!uunet!ukma!darwin.sura.net!sgiblab!munnari.oz.au!metro!extro.ucc.su.OZ.AU!maxtal From: maxtal@extro.ucc.su.OZ.AU (John MAX Skaller) Subject: Re: Zero-length structures and pointer comparisons Message-ID: <1992Oct23.064817.2648@ucc.su.OZ.AU> Sender: news@ucc.su.OZ.AU Nntp-Posting-Host: extro.ucc.su.oz.au Organization: MAXTAL P/L C/- University Computing Centre, Sydney References: <23937@alice.att.com> <1992Oct22.180951.24111@jyu.fi> <23978@alice.att.com> Date: Fri, 23 Oct 1992 06:48:17 GMT Lines: 44 In article <23978@alice.att.com> ark@alice.UUCP () writes: >In article <1992Oct22.180951.24111@jyu.fi> sakkinen@jyu.fi (Markku Sakkinen) writes: > >> Thus, while Jim Adcock, in the posting to which you had originally replied, >> was quite happy with the situation as is, you would be in favour of a change >> in the standard to make pointer types totally ordered? > >Yes, I would. Of course, my preference may not count for all that much, >given the number of machines out there with segmented architectures. >It would be nice if someone with access to real data could do a >cost/benefit analysis. I dont understand how the segmented architectures have anything to do with it. It is only necessary that there be *an* total order, it makes no difference what that order is. Certainly there is no problem on any of the 80x86 machines in obtaining an arbitrary total order on pointers. On the 486 in 32 bit protected mode pointers are 48 bits and can just be compared bitwise just as on a linear address machine. The problems that exist on such machines are not so much with < but with == and !=. In principle on the 486 two machine pointers---different bit patterns---can point to the same object in such a way that it is NOT possible to compare them for equality without invoking the operating system. (This will occur if two segments are overlaid, for example, one being read only and the other read/write. It can also occur on a linear address machine (or the 486) if the virtual memory systemn allows remapping of the linear address space. Here two different linear addresses might point to the same object) Without placing constraints on architectures the best that can be said is that if the bit patterns of two pointers are equal, they point to the same object. If they are not equal, nothing further is known. Even *this* is not true if the pointers are from two different tasks. -- ;---------------------------------------------------------------------- JOHN (MAX) SKALLER, maxtal@extro.ucc.su.oz.au Maxtal Pty Ltd, 6 MacKay St ASHFIELD, NSW 2131, AUSTRALIA ;--------------- SCIENTIFIC AND ENGINEERING SOFTWARE ------------------