lclint-interest message 99

From evans@cs.virginia.edu Wed Sep  4 19:16:09 1996
Date: Wed, 4 Sep 96 16:58:33 -0400
From: evans@cs.virginia.edu (David Evans)
To: lclint-interest@larch.lcs.mit.edu
Cc: ralhv@aals17.alcatel.com.au
In-Reply-To: Vinod Ralh's message of Wed, 04 Sep 1996 08:34:10 -0500 (EST) <9609032234.AA27937@aals7.alcatel.com.au>
Subject: Subscription request - introducing myself!


Hi Vin,

> We're going to initially use lclint as a decent lint tool. After that,
> we'll see how it goes.... We're going to use Purify to catch dynamic
> memory problems and we'll be evaluating Cantata (from IPL) as a testing
> tool.

If you do employ lclint's memory checking also, it'll be interesting to
see what kinds of errors are more effectively detected at run-time vs.
compile-time.  

> As for lclint - how about doing something like Meyers book for Code Wizard
> but for C. There's a classic book about the Traps and Pitfalls of C by
> Koenig - I'm sure it would be very difficult to implement but a great
> challenge!

Some of the checks done by LCLint are indirectly inspired by the "Traps
and Pitfalls" book (in particular the macro checking), and some others
are from Van der Linden, "Expert C Programming: Deep C Secrets" book.  I
don't have a copy of "Traps and Pitfalls" handy, but my recollection is
that some of the problems are things that were corrected by the ANSI
standard.

> Also, what about introducing metrics such as code complexity measurements
> into lclint. I realise that they must be taken with a pinch of salt but
> can be good indicators of dodgey practices.

Yes, adding metrics to lclint has been suggested by a number of people.
I presume some of the standard metrics could be added quite easily
(although I have no plans to do this --- but the source code is
available if anyone wants to give it a try).  More interesting would be
to consider devising new kinds of metrics that exploit the extra
information available to lclint (e.g., abstraction barriers, shared
pointers, clean interfaces), although I don't have any specific ideas as
to what would be useful here in a metric.

--- Dave