Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84 SMI; site sun.uucp
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!genrad!decvax!decwrl!sun!guy
From: guy@sun.uucp (Guy Harris)
Newsgroups: net.micro,net.lang
Subject: Re: Unchecked conversion in Ada
Message-ID: <2323@sun.uucp>
Date: Thu, 20-Jun-85 01:07:45 EDT
Article-I.D.: sun.2323
Posted: Thu Jun 20 01:07:45 1985
Date-Received: Sat, 22-Jun-85 23:46:52 EDT
References: <1649@reed.UUCP> <186@harvard.ARPA> <453@calmasd.UUCP> <192@harvard.ARPA> <266@sdcc13.UUCP> <197@harvard.ARPA>
Organization: Sun Microsystems, Inc.
Lines: 19
Xref: watmath net.micro:10833 net.lang:1624

> a function which converts between any two types bit-to-bit; something
> like C's casts.

C casts are not bit-for-bit conversions unless the conversion happens to be
implemented as such.  This is the case for signed <-> unsigned conversions
on two's complement machines, pointer <-> integer conversions, and pointer
<-> pointer conversions in most cases on byte-addressible machines.

C compilers don't complain about conversions between pointers of different
types, nor do they complain about various other type conversions.  "lint"
will complain about more of them.  I've found bugs in kernel code I've
written that would have been found by "lint" (for those of you who think
strong type checking is a Bad Thing in system code).

Note that the idea of bit-for-bit conversion operators dates back to PL/I's
"unspec" (which, like PL/I's pointers, was typeless); Mesa has a bit-for-bit
type conversion operator "loophole" similar in concept to Ada's.

	Guy Harris