Xref: utzoo comp.lang.c:35138 comp.std.c:4109
Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!usc!samsung!uunet!mcsun!unido!mikros!mwtech!martin
From: martin@mwtech.UUCP (Martin Weitzel)
Newsgroups: comp.lang.c,comp.std.c
Subject: Re: (void *) to function pointer
Message-ID: <1037@mwtech.UUCP>
Date: 10 Jan 91 13:09:55 GMT
References: <1991Jan9.000809.9152@brolga.cc.uq.oz.au>
Reply-To: martin@mwtech.UUCP (Martin Weitzel)
Organization: MIKROS Systemware, Darmstadt/W-Germany
Lines: 46

In article <1991Jan9.000809.9152@brolga.cc.uq.oz.au> gottsc@citrus.citr.uq.oz.au (John Gottsc) writes:
>
>Hello, 
>
>can anybody tell me what the ANSI C standards state regarding conversion
>between a (void *) and a function pointer (i.e. is this type of
>conversion possible in ANSI C)? 

No (as Doug Gwyn allready wrote).

If you need a common storage place for data- and function pointers
you have to use a union. If your function pointers are of different
types (including prototype declarations of function parameters!) you
either must specify all function types as components of the union,
or use a cast before calling the function through the pointer.

As there are no such things as "dynamic casts", i.e. all the casts
are nailed down in the source, this normally means that there must
be somewhere else some information which lets you determine the type
of the function pointer at runtime and then you use a switch statement.

This works because ANSI C requires that "all funtion pointers smell the
same" - but only wrt their storage requirements. There may different
subroutine linkage be used depending on the types of parameters, hence
the correct cast is mandatory for portability.

A less portable solution to transfer a function pointer to a `void *'
is an intermediate cast to an `int'. The ANSI standard allows such
casts but imposes no requirements% on the used mapping functions. So
this may be used as a kluge if you only want to circumvent the
limitation that a function pointer ca not be casted into a `void *'
but don't expect this to work on an other hardware or even on the
same hardware with some other memory modell or some other vendors
compiler or the next release of your compiler. (Don't say I didn't
warn you :-))

--------
%: A footnote in the relevant section of the standard (3.3.4) and some
remarks in the rationale part can be read that it is "a quality of
implementation issue" whether pointer1 -> integral type -> pointer2
result in pointer1 == pointer2. On segmented machine architectures it
may further be possible, that *pointer1 and *pointer2 refer to the
same object but do not compare equal. (All this is true for function
pointers AND for data pointers.)
-- 
Martin Weitzel, email: martin@mwtech.UUCP, voice: 49-(0)6151-6 56 83