Question

It’s been a long (long…) time since I’ve done any C programming and I’m stuck on what should be a really simple issue. I have a simple function that’s calling another function that is allocating some memory on the heap, populating it with a structure and returning a pointer back. Everything looks great, except when the pointer is returned to the calling function, it’s address is corrupted.

Here’s the relevant code:

Called function:

struct server_type *init_gwserver(Cfg* cfg)
{
    struct server_type *res = NULL;
    // Lots of irrelevant code removed here - doesn't reference res.

    res = gw_malloc(sizeof(struct server_type));
    gw_assert(res != NULL);

    res->server_start = gwserver_start; // Pointer to a function
    res->server_stop = gwserver_stop; // Pointer to another function
    return res;
}

Calling function:

struct server_type *do_init_server(Cfg *cfg)
{
    struct server_type *res = NULL;
    res = (struct server_type *)init_gwserver(cfg);
    if (res) {
        return res;
    }
}

OK, here’s the weirdness. When “res” is returned to the calling function, the address the pointer is pointing to changes, and becomes an invalid memory reference. Here’s a quick run through with GDB. Comments (//…) are mine…

Breakpoint 1, init_gwserver (cfg=0x100214e30) at init_gwserver.c:339
339 res->server_stop = gwserver_stop;
(gdb) print /x res
$18 = 0x100215b10
// Pointer above looks fine and seems to be a valid address
(gdb) p *res
$14 = {
    type = 0x100215460, 
    sql_enter = 0x100003820 <gwserver_start>, 
    sql_leave = 0x100005ae0 <gwserver_stop>, 
}
// Contents of the pointer are looking great
(gdb) next
340  return res;
(gdb) print /x res
$19 = 0x100215b10
// Pointer is unchanged - should return this value just fine.
(gdb) next
do_init_server (cfg=0x100214e30) at init_server.c:52
52 if (res) {
(gdb) print /x res
$20 = 0x215b10
// WOW - what the hell happened to the address pointed to by res? It lost the leading 100.
(gdb) p *res
Cannot access memory at address 0x215b10
// Yep, it's definitely now an invalid pointer. Crash time...`

This is probably something really, really simple, but I can’t put my finger on it. Anyone got suggestions?

Thanks!
Toby.

Answer 1

This cast looks suspicious: (struct server_type *)init_gwserver(cfg).

If the cast is necessary to make the code compile then it indicates that there is not a correct prototype visible for init_gwserver at the call site.

If there is no prototype visible, the compiler will make the assumption that the function returns an int. When the return value is retrieved as something that it is not, depending on the implementation, its value may be corrupted and when converted back to a pointer information may have been irretrievably lost.

The cast should not be necessary and if removing it causes a compile error it indicates that you need to add (or #include) a correct prototype for init_gwserver before the point where you attempt to call it.