NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/lang/c/18741

Text File | 1992-12-21 | 1.6 KB | 44 lines

Newsgroups: comp.lang.c
Path: sparky!uunet!taumet!steve
From: [e-mail] (Steve Clamage)
Subject: Re: assert
Message-ID: <[e-mail]>
Organization: TauMetric Corporation
References: <[e-mail]>
Date: Mon, 21 Dec 1992 16:30:06 GMT
Lines: 33

[e-mail] (Hallvard B Furuseth) writes:

>assert is sometimes (ie sun or gcc) defined to call another macro (ie
>__assert) which does most or all of the work.  I expect it's a hack to
>make sure the argument is stringified properly, or something like that.
>Can somebody explain?
>--

It is probably invoking a function rather than another macro.

"assert" has to print output on standard error, generally formatted
in some nice way including the file and line number where the
insertion failed as well as the failed assertion.  The easiest way
to do that is with fprintf().

That in turn would mean that the assert header would have to include
<stdio.h> to get the proper declarations, or duplicate the relevent
declarations from <stdio.h>.

The first alternative (#include <stdio.h>) is forbidden by the C
Standard (standard headers may not #include other standard headers).

The second is messy, and may cause interactions with user code.
(It would impinge on your right to use identifiers from <stdio.h>
when you don't include <stdio.h> in your code.)

Probably the best solution is for the assert macro to call a function
which is compiled separately and can therefore include whatever headers
it needs.  This function is often called something like __assert() or
__gripe(), names which are reserved for the implementation.
--

Steve Clamage, TauMetric Corp, [e-mail]