Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uunet!microsoft!jimad
From: jimad@microsoft.UUCP (Jim ADCOCK)
Newsgroups: comp.lang.c++
Subject: Re: Comparison functions for qsort() and bsearch()
Keywords: Sun, C++
Message-ID: <70271@microsoft.UUCP>
Date: 28 Jan 91 21:32:08 GMT
References: <1990Dec26.160636.15566@clear.com> <1991Jan6.050801.20821@cbnewsk.att.com> <1991Jan9.161633.21922@clear.com> <1991Jan21.022720.20707@cbnewsk.att.com>
Reply-To: jimad@microsoft.UUCP (Jim ADCOCK)
Organization: Microsoft Corp., Redmond WA
Lines: 20

In article <1991Jan21.022720.20707@cbnewsk.att.com> hansen@pegasus.att.com (Tony L. Hansen) writes:
>I'll repeat:  The key to all of this is that whenever calling qsort(), you
>always have to pass it a function which casts from the const void* back to
>the true pointer that was passed in.  And once you do that, you can then cast
>to any base pointer you want!

I don't doubt that this should work on just about any sane compiler out
there today, _but_ I see nothing in ARM that gives one permission to 
explicitly cast back from a void* back to the original object* type.
What C++ does permit is casting back from a pointer to a   
smaller object.  I don't see anywhere ARM defines that the effective size
of a void object, were such allowed to exist, is smaller than any "real" object.
However, most all systems would make a char the smallest object, so one
_should_ be able to cast to/from a char*.  Thus, I claim if one needs
a generic object pointer, one should use a char*, not a void*.  If
you want to represent a generic address, but not imply an object is 
referred to, then feel free to use a void*.  Since section 5.4 does not
explicitly allow casting back from a void*, but most all compilers allow it,
I claim a cast back from void* is an un-constrained error -- IE a programming
error compilers are not required to diagnose.