Hello all:
I'm using a pointer array(by gtk+ lib) to store the function
pointers (in line 83).
Now. I have a big problem. As I get the element of the pointer array, I
have no idea how to cast void pointer in to my function pointer!!!!
The follows inst is my function pointer.
void (*inst)(core_ptr c, GArray* garray);


66 void* initialize_loadshareobjects(){
67         void *handle;
68         char *error;
69
70         handle = dlopen (SHARE_LIB_PATH, RTLD_NOW);
71
72         if (!handle){
73              fprintf (stderr, "%s\n", dlerror());
74              exit(1);
75         }
76         dlerror();
77         return handle;
78 }
79
80 void load_instruction(void *handle, GPtrArray* insts, gchar*
function_nm){
81         void (*inst)(core_ptr c, GArray* garray);
82         //*(void **) (&inst) = dlsym(handle, function_nm);
83         g_ptr_array_add(insts, dlsym(handle, function_nm));
84 }
85

Thanks for your reading!!

[ Post a follow-up to this message ]

"s88" <dave.tw@gmail.com> writes:

> Hello all:
>      I'm using a pointer array(by gtk+ lib) to store the function
> pointers (in line 83).
> Now. I have a big problem. As I get the element of the pointer array, I
> have no idea how to cast void pointer in to my function pointer!!!!

You don't need to cast void pointers.  A simple, unadorned assignment
will the job perfectly.  In fact, it is the *proper* way to do it.
Casting a void * is wrong, unless you use C++ (but then C++ is often
considered wrong in itself).

--
Måns Rullgård
mru@inprovide.com

[ Post a follow-up to this message ]

s88 wrote:
> Hello all:
>      I'm using a pointer array(by gtk+ lib) to store the function
> pointers (in line 83).
> Now. I have a big problem. As I get the element of the pointer array, I
> have no idea how to cast void pointer in to my function pointer!!!!
> The follows inst is my function pointer.
> void (*inst)(core_ptr c, GArray* garray);
>
>
>      66 void* initialize_loadshareobjects(){
>      67         void *handle;
>      68         char *error;
>      69
>      70         handle = dlopen (SHARE_LIB_PATH, RTLD_NOW);
>      71
>      72         if (!handle){
>      73              fprintf (stderr, "%s\n", dlerror());
>      74              exit(1);
>      75         }
>      76         dlerror();
>      77         return handle;
>      78 }
>      79
>      80 void load_instruction(void *handle, GPtrArray* insts, gchar*
> function_nm){
>      81         void (*inst)(core_ptr c, GArray* garray);
>      82         //*(void **) (&inst) = dlsym(handle, function_nm);
>      83         g_ptr_array_add(insts, dlsym(handle, function_nm));
>      84 }
>      85
>
> Thanks for your reading!!
>

So what's wrong with the following:

void (*xxx)(core_ptr, GArray*);
xxx = (void(*)(core_ptr, GArray*)) dlsym(handle, function_nm);

After all void(*)(core_ptr, GArray*) reads as: "an unnamed pointer to a
function returning void with an argument list of (core_ptr, GArray*)".

--

Fletcher Glenn

[ Post a follow-up to this message ]

Fletcher Glenn <fletcher@removethisfoglight.com> writes:

> So what's wrong with the following:
>
> void (*xxx)(core_ptr, GArray*);
> xxx = (void(*)(core_ptr, GArray*)) dlsym(handle, function_nm);

dlsym() returns a void *, so there is no need for a cast.  Needlessly
casting things can hide serious problems that would otherwise have
triggered a compiler warning.

--
Måns Rullgård
mru@inprovide.com

[ Post a follow-up to this message ]

s88 wrote:
> Hello all:
>      I'm using a pointer array(by gtk+ lib) to store the function
> pointers (in line 83).
> Now. I have a big problem. As I get the element of the pointer array, I
> have no idea how to cast void pointer in to my function pointer!!!!
> The follows inst is my function pointer.
> void (*inst)(core_ptr c, GArray* garray);

An illuminating discussion:
http://groups.google.com/group/comp...375c1a6ec2566c1

[ Post a follow-up to this message ]

Måns Rullgård wrote:
> Fletcher Glenn <fletcher@removethisfoglight.com> writes:
>
> 
>
>
> dlsym() returns a void *, so there is no need for a cast.  Needlessly
> casting things can hide serious problems that would otherwise have
> triggered a compiler warning.
>

I guess that I have really gotten used to C++ where not casting
a void * is impossible.

--

Fletcher Glenn

[ Post a follow-up to this message ]

Måns Rullgård <mru@inprovide.com> writes:

> Fletcher Glenn <fletcher@removethisfoglight.com> writes:
> 
>
> dlsym() returns a void *, so there is no need for a cast.  Needlessly
> casting things can hide serious problems that would otherwise have
> triggered a compiler warning.

Generally, this is very good advice. However, in this case it's
wrong. void* can only be implicitly converted to pointers to
incomplete or object type. In this case, the type is a pointer to
function, and requires a diagnostic.

gcc won't complain, though, unless you give it the -pedantic flag.

For maximum portability, though, the cast is a good idea.

Note that C doesn't guarantee that casting it is even well-defined;
but SUS implies it through its definition of dlsym(). It also
specifically calls out this problem in its Rationale (which seems to
have a bug or two).

[ Post a follow-up to this message ]

On 2006-03-13, Måns Rullgård <mru@inprovide.com> wrote:
> Fletcher Glenn <fletcher@removethisfoglight.com> writes:
> 
>
> dlsym() returns a void *, so there is no need for a cast.  Needlessly
> casting things can hide serious problems that would otherwise have
> triggered a compiler warning.

Well, yes, except... well, it's been explained by others in this thread.
This just goes to show that you shouldn't just automatically post
boilerplate answers/criticism like this without a reason. dlsym is not
malloc, its return value is more likely than not to require a cast,
since it's most often used, I suspect, for functions.

[ Post a follow-up to this message ]

Jordan Abel <random832@gmail.com> writes:

> On 2006-03-13, Måns Rullgård <mru@inprovide.com> wrote: 
>
> Well, yes, except... well, it's been explained by others in this thread.
> This just goes to show that you shouldn't just automatically post
> boilerplate answers/criticism like this without a reason. dlsym is not
> malloc, its return value is more likely than not to require a cast,
> since it's most often used, I suspect, for functions.

I stand corrected.

--
Måns Rullgård
mru@inprovide.com

[ Post a follow-up to this message ]