Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uunet!lupine!rfg
From: rfg@NCD.COM (Ron Guilmette)
Newsgroups: comp.std.c++
Subject: Re: "packed" objects
Message-ID: <1312@lupine.NCD.COM>
Date: 24 Aug 90 05:36:18 GMT
References: <56637@microsoft.UUCP> <1227@lupine.NCD.COM> <56735@microsoft.UUCP>
Organization: Network Computing Devices, Inc., Mt. View, CA
Lines: 37

In article <56735@microsoft.UUCP> jimad@microsoft.UUCP (Jim ADCOCK) writes:
>In article <1227@lupine.NCD.COM> rfg@NCD.COM (Ron Guilmette) writes:
>|
>|The problem of packing across inheritance boundaries is a different
>|matter however...
>
>Okay, sounds interesting.  How about a proposal about how this should
>work?  I'd like to propose that your suggestion should still allow
>portability across machines with differing packing strategies, and should
>not be onerous to people who aren't interested in these details....

I could have sworn that I already *did* make a proposal.  How about:

	class base { int i; };
	class derived : public packed base { int j; };

I'm assuming that the programmer might wish to specify exactly *which*
base class is supposed to get packed up against the fields of the derived
class.

Of course the example given above will never fly (politically) because it
introduces a new keyword.  (God forbid :-)  Of course you could always get
around that problem in the standard C++ way, i.e. by reusing yet another
existing keyword which should not be reused and by assigning it some new
and different meaning in one particular syntactic location.  So how about:

	class base { int i; };
	class derived : public auto base { int j; };

It reads poorly, and it the meaning has nothing whatsoever to do with
auto'ness, but hey!  It creates no new keywords! :-)

-- 

// Ron Guilmette  -  C++ Entomologist
// Internet: rfg@ncd.com      uucp: ...uunet!lupine!rfg
// Motto:  If it sticks, force it.  If it breaks, it needed replacing anyway.