lclint-interest message 73

From evans@cs.virginia.edu Wed May  8 15:16:49 1996
Date: Wed, 8 May 96 15:11:05 -0400
From: evans@cs.virginia.edu (David Evans)
To: lclint-interest@larch.lcs.mit.edu
Cc: bill@ellie.etec.com
In-Reply-To: Bill Margolis's message of Tue, 7 May 96 17:08:17 PDT <9605080008.AA23062@mailboy>
Subject: Re: Conformance checking capabilities?


> However, a useful, immediate task that we are currently facing is
> to see how legacy and currently developing source code could be
> checked against the various posix, svid 3(whatever that is), and
> other standards of orthodoxy.  I think the usual type of messages with 
> hints that are currently in lclint is the kind of aid that we would 
> want to accept.  (People in our group have been visited by the 
> electronic fairies from Abraxas and MKS Code Integrity).
>
> Are there libraries that do these things on the agenda?  
> Perhaps they are already available, if only I were to key in the proper
> netscape location, or make and install the right library?

It would be very useful to have libraries like this, but as far as I
know they don't exist and we have no plans (or resources) to create
them.  Currently, only the ANSI Standard library (two versions,
depending on the strictness of checking desired) and a UNIX library
(based on what is available on *most* unixes) included in the lclint
distribution are available.  I hope lclint users will provide annotated
versions of various standards and system header files, and I'll add
these to the lclint releases if they become available.

Adding appropriate lclint annotations to the header files for a given
standard should be fairly painless, as long as reasonable header files
and documentation can be found.  I'm not sure about the posix and svid 3
standards, but a major difficulty in creating the ANSI standard version
of the library is that the documentation (in this case, the ANSI
Standard) is often very unclear about interfaces for these functions
(e.g., which pointer parameters can be null, does the return value point
to storage that must be deallocated or is it a local buffer, what
parameters are valid).  One hopes eventually standards will be
documented in a more precise way, but I'm not counting on this happening
anytime soon.

--- Dave