NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/std/c/768

Internet Message Format | 1996-08-06 | 2KB

Path: xylogics.com!not-for-mail
From: [e-mail] (James Carlson)
Newsgroups: comp.std.c
Subject: Re: Q: char **foo, char *foo[], and char foo[][] ?
Date: 19 Apr 1996 08:20:11 -0400
Organization: Xylogics Incorporated
Distribution: world
Message-ID: <4l80dr$[e-mail]>
References: <4l33ok$[e-mail]> <[e-mail]>
Reply-To: [e-mail]
NNTP-Posting-Host: newhub.xylogics.com


In article <[e-mail]>, [e-mail] (J. Kanze) writes:
|> In article <4l33ok$[e-mail]> [e-mail] (Philippe
|> Maurisset) writes:
[...]
|>
|> |> myexample()
|> |> {
|> |>     char foo[MAX_X][MAX_Y];
|> |>     ...
|> |>     myfunc( (char **)foo );
|> |> }
|>
|> |> void myfunc( char *foo[MAX_X] )
|>                 ^^^^^^^^^^^^^^^^
|>
|> Because it appears as a function parameter, this is actually a
|> declaration of a char**, and not a char*[].

True.

|> So in fact, you have no type incompatibility to deal with.

Not true.  The original definition of foo in myexample() was as a two-
dimensional array, which, in C, is implemented in storage as a
unidimensional array of size MAX_X*MAX_Y*sizeof(char), not as an array
of pointers-to-characters.  If you try to access this as a pointer-to-a-
pointer, you will be attempting to dereference character data as a
pointer.  (If you've got a real operating system, then this just results
in a core fault.  If you're running on a toy, the results could be far
worse.)

(Of course, the type of the formal argument is compatible with the
cast given in the call, but the cast itself is bogus.)

--
James Carlson <[e-mail]>            Tel:  [phone]
Annex Interface Development / Xylogics, Inc.          [phone]
53 Third Avenue / Burlington MA  01803-4491     Fax:  [phone]