Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.std.c > #1387

Re: Conversion from void* to pointer-to-function?

Show all headers | View raw

вторник, 26 июня 2012 г., 1:57:34 UTC+4 пользователь Keith Thompson написал:
> James Kuyper <jameskuyper@verizon.net> writes:
> > On 06/25/2012 04:26 PM, Keith Thompson wrote:
> >> Consider the following toy program:
> >> 
> >> int main(void) {   
> >>     void *void_ptr = 0;
> >>     void (*func_ptr)(void) = (void(*)(void))void_ptr;
> >>     return 0;
> >> }
> >> 
> >> (Note that void_ptr is *not* a null pointer constant, it's merely a
> >> pointer object that happens to contain a null pointer.)
> >> 
> >> Is the conversion of void_ptr to void(*)(void) (a pointer-to-function
> >> type) permitted?
> >> 
> >> N1570 6.5.4 lists three constraints for the cast operator. This code
> >> doesn't violate any of them.
> >> 
> >> N1570 6.3.2.3 describes the semantics of conversions involving pointers.
> >> It mentions conversions from void* to and from pointer-to-object types,
> >> conversions from one pointer-to-object type to another, and conversions
> >> from one pointer-to-function type to another.  Conversions from void* to
> >> a pointer-to-function type simply aren't mentioned.
> >
> > Conversion of null pointers to other pointer types is mentioned
> > (6.3.6.3p4). It applies only to null pointers being converted to pointer
> > types, but there are no other type-based restrictions on the
> > applicability of that section.
> >
> >> My conclusion is that the conversion is permitted (it doesn't violate
> >> any constraint or syntax rule), but its behavior is undefined by
> >> omission.
> >
> > The behavior is defined by 6.3.6.3p4: it produces a null pointer of type
> > void(*)(void).
> 
> It's 6.3.2.3p4, not 6.3.6.3p4.
> 
> You're right, of course -- which makes me realize I didn't ask what I
> intended to ask.  I meant to ask about conversion from void* to a
> pointer-to-function type in general, not about conversion of null
> pointers.
> 
> A revised test case:
> 
> #include <stdlib.h>
> #include <assert.h>
> int main(void) {   
>     void *void_ptr = malloc(1024);
>     assert(void_ptr != NULL);
>     void (*func_ptr)(void) = (void(*)(void))void_ptr;
>     return 0;
> }
> 
> Assume that the assert doesn't fire.
> 
> I believe that the conversion in the initializer for void_ptr is
> legal (i.e., doesn't violate any syntax rule or constraint) but has
> undefined behavior by omission.  (The authors of gcc, based on the
> warning given with "gcc -std=c99 -pedantic", apparently disagree.)
> 
> Is my interpretation correct?
> 
> I know that the standard explicitly allows for undefined behavior when
> the behavior is quietly not specified, but IMHO the standard would be
> improved if 6.3.2.3 covered this case and explicitly stated that the
> behavior is undefined.
> 
> -- 
> Keith Thompson (The_Other_Keith) kst-u@mib.org  <http://www.ghoti.net/~kst>
>     Will write code for food.
> "We must do something.  This is something.  Therefore, we must do this."
>     -- Antony Jay and Jonathan Lynn, "Yes Minister"

I think there is another interpretation of this issue.

6.2.5 p.20: "A pointer type may be derived from a function type or an object type, called the referenced type".
"referenced type" is highlighted and supposed to be important.

6.3.2.1 p.8: "If a converted pointer is used to call a function whose type is not compatible with the referenced type, the behavior is undefined."

So, this statement should be applied to both object and function pointers and not only to function pointers. 6.2.5 states that there are two groups of pointers: pointers-to-object and pointers-to function. It doesn't state that there is a third case of void* pointer, so void* should be pointer-to-object because its type is incomplete (object types may be incomplete) and it may be freely converted to any pointer to object.

Back to comp.std.c | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

Conversion from void* to pointer-to-function? Keith Thompson <kst-u@mib.org> - 2012-06-25 13:26 -0700
  Re: Conversion from void* to pointer-to-function? James Kuyper <jameskuyper@verizon.net> - 2012-06-25 17:23 -0400
    Re: Conversion from void* to pointer-to-function? Keith Thompson <kst-u@mib.org> - 2012-06-25 14:57 -0700
      Re: Conversion from void* to pointer-to-function? Maxim Fomin <maxim@maxim-fomin.ru> - 2012-06-27 10:03 -0700
        Re: Conversion from void* to pointer-to-function? Keith Thompson <kst-u@mib.org> - 2012-06-27 11:33 -0700
          Re: Conversion from void* to pointer-to-function? David R Tribble <david@tribble.com> - 2012-07-24 20:05 -0700
  Re: Conversion from void* to pointer-to-function? Tim Rentsch <txr@alumni.caltech.edu> - 2012-12-21 00:32 -0800
    Re: Conversion from void* to pointer-to-function? Tijl Coosemans <tijl@coosemans.org> - 2012-12-21 10:00 +0100
    Re: Conversion from void* to pointer-to-function? Keith Thompson <kst-u@mib.org> - 2012-12-21 02:01 -0800
      Re: Conversion from void* to pointer-to-function? Tim Rentsch <txr@alumni.caltech.edu> - 2012-12-21 08:12 -0800
        Re: Conversion from void* to pointer-to-function? Keith Thompson <kst-u@mib.org> - 2012-12-21 17:58 -0800
          Re: Conversion from void* to pointer-to-function? Tim Rentsch <txr@alumni.caltech.edu> - 2012-12-22 08:22 -0800

csiph-web