NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part2/10868

Internet Message Format | 1996-08-05 | 3.5 KB

Path: binah.cc.brandeis.edu!DERNIS
From: [e-mail]
Newsgroups: comp.lang.ada,comp.lang.c,comp.lang.c++
Subject: Re: C/C++ knocks the crap out of Ada
Date: 11 Mar 1996 02:01:15 GMT
Organization: Brandeis University
Message-ID: <4i01hb$[e-mail]>
References: <[e-mail]> <[e-mail]>    <[e-mail]> <4h5hgj$[e-mail]>    <[e-mail]> <[e-mail]> <[e-mail]>
Reply-To: [e-mail]
NNTP-Posting-Host: binah.cc.brandeis.edu

Sorry I lost the credit:
>
>To be precise, C standard declares all programs that violate array
>bounds as illegal, but does not _require_ the compiler to check for
>these violations. This is a change from pre-standard days when a
>program was considered valid as long as it did not try to access
>memory beyond what was allocated to it.
>
>As an example,
>
>char a2[100][100];
>char *a = &a2[0][0];
>
>It is obvious from the constraints of memory layout for arrays and
>pointer arithmetic, that a+200 would have pointed to a[2][0] if it
>were valid. However, the standard has declared a+200 to invoke
>undefined behaviour. Thus, a very strict bounds checking
>implementation can reject such pointer arithmetic without violating
>the standard.

It does little good to tell a programmer not to do
something in the language standard when I bet most
programmers have never read the ARM, and certainly
do not keep up with the latest version.


I have jus read a good deal of this thread.  Several people
have made this argument to me that many languages not just
ADA impose more (better?) structure, on the programmer.  Thus
resulting in better reliability.  It is also readily agreed that
this is done at some expense in terms of memory requirments,
and also speed (especially in tight loops).

I went from the evolution from C to C++,  and feel that many
of these problems can be easily solved if programmers just
took advantage of the new features.  C++ can be safer.  For
example, if you write (or buy) a good array class and use it
instead of the standard types, then you will know longer have
run away arrays.  Additionally, you can perform range checking
or anything else.  For that matter, a heap class can be defined
to handle garbage collection as well. Some have set let's compare
languages as they are.  I recall when I read Lipmann's book
that much of the addition of C++ was so that encapsulation could
save the world from sloppy code.  So it is *fair* in my mind
to say that on large projects which are inherently difficult to
debug,  bound checking class be used de rigor.  However, in a tight loop
you are free to use the standard types with no range or bounds
checking.  Just be damn sure it is contained in a small block
of code and that you don't screw it up.

 It would seem to me
that any language which incorporates range checking, bound
checking, garbage collection, etc. from the get go loses the
flexibility of higher performance in critical areas.  This may
not be a problem for most projects since memory has become
so cheap and computers so fast, but many people, particularly
poor graduate students like me, would like their current
hardrives and CPU RAM to hold out for a while.

I am always a few months/years behind on standards.  Are there
standard container classes?  I think the standard "c" libraries,
streams, and complex libraries, are not nearly enough.  Well
implimented standard libraries for safer programming seems like
a logical direction of evolution.

Mitch Dernis