Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!titan!drack
From: drack@titan.tsd.arlut.utexas.edu (Dave Rackley)
Newsgroups: comp.lang.c
Subject: Re: Should #include <foo.h>; compile?
Message-ID: <DRACK.91Feb19085423@diablo.titan.tsd.arlut.utexas.edu>
Date: 19 Feb 91 14:54:23 GMT
References: <7605@jhunix.HCF.JHU.EDU>
Sender: news@titan.tsd.arlut.utexas.edu
Distribution: usa
Organization: Applied Research Laboratories, University of Texas at Austin.
Lines: 36
In-reply-to: barrett@jhunix.HCF.JHU.EDU's message of 19 Feb 91 03:28:01 GMT

In article <7605@jhunix.HCF.JHU.EDU> barrett@jhunix.HCF.JHU.EDU (Dan Barrett) writes:

>   Should this program compile?
>
>       #include <stdio.h>;
>       main() {}

>   I have one compiler that complains about the semicolon after the ">", and
>   one that doesn't.  Which is right?  My suspicion is that the program is
>   legal, because a semicolon by itself is an empty statement.  The error
>   message produced by the picky compiler is:
>
>       test.c:1: #include expects "fname" or <fname>

>   Both compilers accept:
>
>       #include <stdio.h>
>       ;
>       main() {}
>

Preprocessor directives are bounded by *line boundaries* not semi-colons.
Per K&R, 2d edition, page 228:  "...any character not otherwise defined is 
taken as a token."  Therefore, it is my opinion that your first example should
not compile (precompile).

--

  DISCLAIMER:  I, hereby, disclaim any disclaimer except this disclaimer.  

+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=+
| David Rackley                 | Now comes the really weird part...You know  |
| Applied Research Laboratories | the part where I know how to tap dance, but |
| The University of Texas       | only while wearing golf shoes...            |
| Austin, TX.  78758            |        ...Ponderous, man, really ponderous! |
+=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=+