NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c-part1/4303

Internet Message Format | 1996-08-05 | 3.0 KB

Path: Rezonet.net!news
From: [e-mail] (Ray Dunn)
Newsgroups: comp.lang.c
Subject: Re: Please help ?!
Date: 31 Jan 1996 17:25:27 GMT
Organization: Ultimate Technographics Inc.
Message-ID: <4eo8m7$[e-mail]>
References: <4dm889$[e-mail]> <4drnv1$[e-mail]> <4drq5i$[e-mail]> <[e-mail]>
NNTP-Posting-Host: 204.19.230.7
Mime-Version: 1.0
Content-Type: Text/Plain; charset=US-ASCII
X-Newsreader: WinVN 0.99.7

In referenced article, Tanmoy Bhattacharya says...
>This issue keeps coming up in this newsgroup, and as any stylistic
>question, will probably never be `finally answered'. On compilers
>which warn on undeclared functions, avoiding the cast has little
>benefit in this particular case. One must keep in mind, however, that
>programming is more a matter of habit: and, often, it is best not to
>become too used to facilities one has at hand on a particular day on a
>particular platform. The compiler _need not_ warn about an undeclared
>function (i.e. such a warning is not mandated by ANSI: no warning is
>however ever prohibited). Converting an int to a pointer without a
>cast, however, needs a warning. On such compilers, then, avoiding the
>cast can often help spot a fatal error.

Perhaps I'm falling into the Trap that Tanmoy points out, i.e. relying
on my tools to give warnings under these circumstances, but I feel very
strongly that if your compiler does not point out these possible
errors, then some version of lint should be run on code as part of a
consistency check.

Perhaps the use of such a facility belongs in Peter's recent rules for
posting in c.l.c.

>The point that Ray brings up is also one discussed here quite
>often. The first thing is that writing
>
>long_ptr = malloc (sizeof (int));
>
>is inherently more error prone than writing
>
>long_ptr = malloc (sizeof*long_ptr);
>
>This second form is useful to learn as an idiom, and avoids most of
>the errors that the explicit cast is trying to flag.

I agree that this is the best solution.

>A related point is that a `cast' is a loophole in type-safety, just as
>a `goto' is a loophole in structured programming. Both have their
>legitimate uses, and in the hands of an expert programmer, both of
>them can integrate into magnificient products of art. A beginner
>however is often advised to stay away from a goto, I wish the same
>were done for casts. Unfortunately, C does not allow complete
>elimination of casts: but avoiding them whenever possible is probably
>an important lesson that every beginner needs to have.

Unfortunately, many uses of "void *" are far bigger loopholes in type
safety than a cast.  Compilers or lint will pick up most erroneous uses
of a cast.  A wrong pointer type as the value of a void * function
parameter can only be debugged at runtime - [a pox on MS for their use
of general handle manipulating functions with void * parameters (:-(].

--
Ray Dunn (opinions are my own) | Phone: [phone]
Montreal                       | Phax : [phone]
[e-mail]          | Home : [phone]