Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!linus!decvax!decwrl!sun!jfarrell
From: jfarrell@sun.UUCP (Jerry Farrell)
Newsgroups: net.lang.c
Subject: Re: type-checking function arguments
Message-ID: <313@sun.UUCP>
Date: Wed, 14-Sep-83 13:41:36 EDT
Article-I.D.: sun.313
Posted: Wed Sep 14 13:41:36 1983
Date-Received: Wed, 14-Sep-83 19:22:38 EDT
References: bronze.747
Lines: 16

I want the compiler to type-check for me because
  1) I'm tired of being handed a program with the remark "OK, it's
	fully tested and solid." and discovering 6 places where foo is 
	passed when it should be *foo;
  2) Lint barfs on constructions the compiler accepts (e.g. /**/ ina
	macro definition, but it doesn't get fixed, because we can creep
	along without it, and there's all these other things . . . .
   3) Preparing a lint library is a royal pain -- unless I'm willing to
	reconstruct cc's lexical analyzer (including keeping the comments),
	I get to massage my source by hand into the appropriate format;
   4)  Probably as a result of (3), the lint libraries tend to be about 3
	years behind the code.
Compilers need not have trouble with any of the difficulties listed in Lantz'
message;  see Mesa, for instance.  The claim that lint's optionality saves
you from paying compile overhead all the time while still providing safety
needs a lot more empirical backing than I've seen.