Xref: utzoo comp.lang.c:24742 comp.lang.c++:5979 comp.std.c:2297
Path: utzoo!utgpu!jarvis.csri.toronto.edu!clyde.concordia.ca!uunet!aplcen!haven!udel!rochester!quiroz
From: quiroz@cs.rochester.edu (Cesar Quiroz)
Newsgroups: comp.lang.c,comp.lang.c++,comp.std.c
Subject: Re: So where is a lint for ANSI C code?
Summary: Better error messages in single compilations are not enough
Message-ID: <1989Dec27.035411.23602@cs.rochester.edu>
Date: 27 Dec 89 03:54:11 GMT
References: <16355@megaron.cs.arizona.edu> <GUMBY.89Dec24124800@Gang-of-Four.Stanford.EDU>
Reply-To: quiroz@cs.rochester.edu (Cesar Quiroz)
Followup-To: comp.lang.c
Organization: University of Rochester, Department of Computer Science
Lines: 43

gumby@Gang-of-Four.Stanford.EDU (David Vinayak Wallace) wrote:
| But GCC and G++ by themselves are pretty good; with -Wall I doubt you
| need a special lint.

Some of lint's finest moments come when one runs it on *several* C
sources simultaneously.  The usual style of Unix C compilation is to
perform separate runs on each file for object code production, but
just one lint run on the whole thing[1].  Admittedly, prototypes
catch one of the more common reasons for having all the files
visible at once (caller-callee disagreement).  Still, a system-wide
checker that is not also expected to generate code could prove
useful.

Anecdote: once I helped some people debug a program that had a weird
case of def-def disagreement.  I think it involved a program that
declared the same structure in two files, altering the order of two
of the fields.  This was just poor coding, of course, but (1) none
of the -c compiles saw anything bad, (2) dbx just went crazy on it.
I recall that I insisted on a full-system lint and saw there the
double, inconsistent, definition.  This was under Sun OS x.y, where
x was most likely something like 2).  I tried reconstructing the
case for posting not long ago, but I failed under Sun OS 4; perhaps
someone more patient can produce an example based on the above
information.

Have a happy New Year!  And let's hope we get a C standard this
year.

Cesar ( Why didn't they invent a trigraph to let me put an accent on
that e?  I thought the C committee was into character set
standardization and transmission protocols too, right? :-)

[1] Unbelievers don't run lint at all.  Naive believers run lint on
each file separately.  None of these strategies is particularly
persuasive, although both yield plenty of excitement during
debugging.


-- 
                                      Cesar Augusto Quiroz Gonzalez
                                      Department of Computer Science
                                      University of Rochester
                                      Rochester,  NY 14627