Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!email!vmars!hp
From: hp@vmars.tuwien.ac.at (Peter Holzer)
Newsgroups: comp.lang.c
Subject: Re: "void **" : Is it valid and what does it mean?
Message-ID: <1991May28.115914.19879@email.tuwien.ac.at>
Date: 28 May 91 11:59:14 GMT
References: <AJ3U.91May27213633@jade.cs.virginia.edu> <6001@goanna.cs.rmit.oz.au>
Sender: news@email.tuwien.ac.at
Distribution: comp
Organization: Technical University of Vienna, Dept. for Realtime Systems
Lines: 49
Nntp-Posting-Host: nowhere.vmars.tuwien.ac.at

ok@goanna.cs.rmit.oz.au (Richard A. O'Keefe) writes:

[...good explanation why void ** isn't a pointer to any pointer
deleted...]

>> Basically what I want is a type that would coerce to a pointer to a
>> pointer to a type, for any type.  Sort of one level beyond simple
>> "void *".

>Why do you want to do _that_?  If you _could_ do it, you would be
>throwing away information that the receiver would need in order to
>be able to _use_ the thing?  What's the context?

I don't know what the original poster wanted to do, but since I was in
the same situation once, I can give you an example why anybody would
like to do something like that.

In MARS (a real-time system developed at our department) exists a
system-call which is now defined as:

int receivem (msg_name name, void * msg, unsigned short flags);

Messages are (in C) defined as structures and there are several types.
For some reason, which is not relevant here, msg is not a pointer to
the structure, but a pointer to a pointer to structure. Of course one
common error is to define a struct foo * mymsg and then pass mymsg
instead of &mymsg. With the current prototype this error is not
detected by the compiler, but if some type `pointer to any pointer'
existed, this type of error could be catched. Unfortunately this is not
possible in C.

#pragma speculation_mode on	/* :-) */
There is a rumor among C-programmers that `all struct-pointers smell
the same.' (I could not find any assertion of this in the standard, and
I can find pro's and con's for such a rule, so I call it a rumor,
undoubtedly someone who knows will tell me if it is correct or not). If
this rumor is true, a pointer to pointer to any struct would be legal
(At least there is no reason why it should not be legal). So we could
declare the systemcall above as:

int receivem (msg_name name, struct ** msg, unsigned short flags);

which is almost exactly what the designer of the system call had in
mind.
--
|    _  | Peter J. Holzer                       | Think of it   |
| |_|_) | Technical University Vienna           | as evolution  |
| | |   | Dept. for Real-Time Systems           | in action!    |
| __/   | hp@vmars.tuwien.ac.at                 |     Tony Rand |