function pointer to itself

[kalculus]

Hello

Can someone explain what the following code is doing?

void *get_gp_value()
{
void **function_poin ter = (void **)get_gp_value ;
return function_pointe r[1];
}

Thanks

[Eric Sosman]

kalculus wrote On 02/17/06 13:01,:

Hello

Can someone explain what the following code is doing?

void *get_gp_value()
{
void **function_poin ter = (void **)get_gp_value ;
return function_pointe r[1];
}

It's some kind of non-portable hack, something
specific to one particular compiler. It seems to be
trying to retrieve the second "word" of the compiled
code of the function. The meaning of that "word" is
not defined by C; it's something to do with the way
functions are compiled on a particular machine. C
doesn't even guarantee that the "word" can be retrieved
at all; on some machines "code" cannot be treated as
"data."

As far as C is concerned, the only thing this code
does is break the rules.

--
Er*********@sun .com

[Keith Thompson]

"kalculus" <su***********@ yahoo.com> writes:

Can someone explain what the following code is doing?

void *get_gp_value()
{
void **function_poin ter = (void **)get_gp_value ;
return function_pointe r[1];
}

Invoking undefined behavior.

HTH, HAND.

--
Keith Thompson (The_Other_Keit h) ks***@mib.org <http://www.ghoti.net/~kst>
San Diego Supercomputer Center <*> <http://users.sdsc.edu/~kst>
We must do something. This is something. Therefore, we must do this.

[John F]

"Keith Thompson" wrote:

"kalculus" wrote:

Can someone explain what the following code is doing?

void *get_gp_value()
{
void **function_poin ter = (void **)get_gp_value ;
return function_pointe r[1];
}

Invoking undefined behavior.

OT:
I wonder how the compiler handles

(void*)[1] ... since sizeof void is not defined...

--
John

kalculus, please email me what type of compiler you are using and an object
file with a void main(...) and the function.
I wanna have a look at it :-)

johnny.f "at" gmx.at

[Chris Torek]

>> "kalculus" wrote:

Can someone explain what the following code is doing?

void *get_gp_value()
{
void **function_poin ter = (void **)get_gp_value ;
return function_pointe r[1];
}

"Keith Thompson" wrote:

Invoking undefined behavior.

(which it is indeed doing - it also returns nothing at all useful
on most machines)

In article <43************ ***********@tun ews.univie.ac.a t>,
John F <sp**@127.0.0.1 > wrote:OT:
I wonder how the compiler handles

(void*)[1] ... since sizeof void is not defined...

This is a syntax error, even in most non-C-but-kind-of-like-C languages.

Assuming you mean:

void *x = <some valid expression>;
... x[1] ...

Standard C requires a diagnostic.

A number of languages that vaguely resemble, but are not in fact, C,
allow this. What they do depends on the language.

Note that gcc by default implements a language almost but not entirely
unlike C :-) and you have to tell it to use "-std=c89" (or "-ansi" in
older versions of gcc), or "-std=c99" for C99, and "-pedantic", to
get it to implement Standard C.

Note that:

void **x = ...
... x[1] ...

is very different from:

void *x = ...
... x[1] ...

The former actually means something in C, provided "x" points to the
first of at least two objects of type "void *". For instance, consider
the following somewhat silly code fragment:

int a;
double b;
struct S { char *p; char *q; } c;

void *three_void_sta rs[3] = { &a, &b, &c };

void **x = three_void_star s;

void *pick(int i) {
return x[i];
}

A call to pick() at this point is valid if it supplies an argument
between 0 and 2 inclusive. The value returned is &a, &b, or &c
(converted to "void *"), correspondingly . Thus, we can finish this
code off with:

#include <stdio.h>

int main(int argc, char **argv) {
int *ip;
double *dp;

ip = pick(0);
*ip = 42;
dp = pick(1);
*dp = 4.2;

if (argc > 1)
printf("pick(1) : %f\n", *(double *)pick(1));
else
printf("pick(0) : %d\n", *(int *)pick(0));
return 0;
}

This program is quite useless, but its output is well-defined.
--
In-Real-Life: Chris Torek, Wind River Systems
Salt Lake City, UT, USA (40°39.22'N, 111°50.29'W) [phone]
email: forget about it http://web.torek.net/torek/index.html
Reading email is like searching for food in the garbage, thanks to spammers.

[tmp123]

John F wrote:

"kalculus" wrote:

Can someone explain what the following code is doing?

void *get_gp_value()
{
void **function_poin ter = (void **)get_gp_value ;
return function_pointe r[1];
}

OT:
I wonder how the compiler handles

OT? pointer usage in C?
(void*)[1] ... since sizeof void is not defined...

sizeof (void *) IS defined.

[John F]

"tmp123" wrote:

John F wrote:

> "kalculus" wrote:
>> Can someone explain what the following code is doing?
>>
>> void *get_gp_value()
>> {
>> void **function_poin ter = (void **)get_gp_value ;
>> return function_pointe r[1];
>> }

OT:
I wonder how the compiler handles

OT? pointer usage in C?

Pointer usage not, but how a compiler does implement it is OT.

I was referring to the use of [] on a void* since it can't be dereferenced
according to my knowledge of the standard. (I actually misread the OPs **,
which is indeed different)

(void*)[1] ... since sizeof void is not defined...

sizeof (void *) IS defined.

I know. Should be sizeof(char*) then. But it is not obliged to fit that.

John

[John F]

"Chris Torek" wrote:

"kalculus" wrote:
Can someone explain what the following code is doing?

void *get_gp_value()
{
void **function_poin ter = (void **)get_gp_value ;
return function_pointe r[1];
}

"Keith Thompson" wrote:

Invoking undefined behavior.

(which it is indeed doing - it also returns nothing at all useful
on most machines)

that's why I asked the OP to email an object file or disassembly on his
machine, that's the only way to tell what it _actually_ does. It is unlikely
to be useful... (I don't see a rationale for that...)
John F wrote:

OT:
I wonder how the compiler handles

(void*)[1] ... since sizeof void is not defined...
This is a syntax error, even in most non-C-but-kind-of-like-C languages.

I didn't mean it litterally (sic!) :-)
Assuming you mean:

void *x = <some valid expression>;
... x[1] ...
indeed.
Standard C requires a diagnostic.

A number of languages that vaguely resemble, but are not in fact, C,
allow this. What they do depends on the language.
Well, yes. It violates the (void*)-not-dereferenceable part of the std.
Note that gcc by default implements a language almost but not entirely
unlike C :-) and you have to tell it to use "-std=c89" (or "-ansi" in
older versions of gcc), or "-std=c99" for C99, and "-pedantic", to
get it to implement Standard C.
As does -a for Open Watcom.
Note that:

void **x = ...
... x[1] ...

is very different from:

void *x = ...
... x[1] ...
I see. I misread the OP (putting the "function pointer"-topic into my
input-filter...). I should go to sleep.
The former actually means something in C, provided "x" points to the
first of at least two objects of type "void *". For instance, consider
the following somewhat silly code fragment:
I see the difference now.

<snip suspicious lines of code>
A call to pick() at this point is valid if it supplies an argument
between 0 and 2 inclusive. The value returned is &a, &b, or &c
(converted to "void *"), correspondingly . Thus, we can finish this
code off with:
<snip>
This program is quite useless, but its output is well-defined.

I agree.

--
John

[Jordan Abel]

On 2006-02-18, John F <sp**@127.0.0.1 > wrote:

I know. Should be sizeof(char*) then. But it is not obliged to fit that.

It's not? As far as i know, it's guaranteed by the standard that char *
and void * have identical representation [same size, same alignment, can
be passed interchangeably to variadic functions or functions with no
prototype, same bit pattern represents the same address]

[John F]

"Jordan Abel" wrote:

On 2006-02-18, John F wrote:

I know. Should be sizeof(char*) then. But it is not obliged to fit that.

It's not? As far as i know, it's guaranteed by the standard that char *
and void * have identical representation [same size, same alignment, can
be passed interchangeably to variadic functions or functions with no
prototype, same bit pattern represents the same address]

Well yes... Just reread... you are right with that!
Sorry for not checking in the first place.

--
John