From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID
	autolearn=no autolearn_force=no version=3.4.4
X-Google-Language: ENGLISH,ASCII-7-bit
X-Google-Thread: 103376,9e2776c05028676e
X-Google-Attributes: gid103376,public
From: dewar@merv.cs.nyu.edu (Robert Dewar)
Subject: Re: Why Ada is not the Commercial Lang of Choice
Date: 1997/06/20
Message-ID: <dewar.866864714@merv>#1/1
X-Deja-AN: 251497469
References: <33A7FBFF.29D2@mitre.org> <5o9eca$aoi$1@goanna.cs.rmit.edu.au>
 <EC03q1.J3A@world.std.com> <33ab1c1c.2926201@news.mhv.net>
Organization: New York University
Newsgroups: comp.lang.ada
Date: 1997-06-20T00:00:00+00:00
List-Id: <comp.lang.ada>


Paul says

<<This is all well and good when you know the exact address to set the
variable to. The problem comes with a variable of type "address" that
you need to perform masking on. I was unable to re-cast a variable of
type "address" (which was derived from unsigned integer) to an
unsigned integer  so that I could mask out the lower 48 bits. That's
why I went the "unchecked conversion" route. Ada seems to be so
strictly typed that even though a variable is typed or sub-typed as
unsigned integer, it is treated as a completely different type than a
variable typed as unsigned integer.>>

Interesting that you choose an example where Ada has preferable semantics
to C. In C there is no way to treat an address as an integer, except to
do the equivalent of an unchecked conversion (i.e. a cast to an integral
type). This cast is tricky, because

(a) you have no way of declaring the integer type to which you will cast
it in a portable manner. In fact it is not even required that there be
such a type, though there usually will be (it might be int, or it might
be long, or it might be long long).

(b) when you do this cast in C, you have no idea what the equivalence will
be, so as soon as you do things like masking out lower order bits you are
in quite implementation dependent territory.


Now obviously no language can guarantee portable semantics for this type
of operation, but Ada does a much better job of providing appropriate
semantic facilities here. In particular there is a defined type
Integer_Address, which is precisely defined as an integer equivalent to
a linear address, if this makes sense for the machine. And one can convert
Address values to and from Integer_Address without using Unchecked_Conversion.

Going the "unchecked conversion" route in Ada 83, which lacks this capability
makes perfect sense, but how on earth could you present this as a difference
between Ada and C. The unchecked conversion from integer to address in Ada
has exactly the same semantics as the (unchecked) cast in C from integer
to address.

I often notice that when you get people to give actual examples in C of
code that they would like to present as demonstrating the inadequacies
of Ada, they actually turn out to be great examples of the superiority
of Ada over C, in the realm which people think of as belonging to C,
namely low level operations.

That's why it is a shame that Joe will not or cannot follow through on
giving specific examples. They very likely would be cases of this kind!