unnecessary casts

[Miika Komu]

DB: There are gazillions of casts, often two in a row. Casts should be
the exception, not the norm. If a large number of casts are
required, then the code structure and design is plain wrong.

Often the results of functions returning void* are cast to some type
even though void* is compatible with any pointer (C is not C++). I
already removed such pointless casts for malloc, memcpy, memset, free
and similar.

But many instances still remain, as an example, take the function
hip_get_param_contents_direct, it returns void*, but its results are
cast all the time, e.g. in lib/conf/conf.c.

MK: Suggest a systematic way to find casts?

---

DB: Another example from lib/opphip/wrap_db.c:

    ret = (hip_opp_socket_t *) hip_ht_find(socketdb, (void *) &opp);

Both casts are unnecessary, this is the prototype for hip_ht_find:

  void *hip_ht_find(void *head, void *data);

My personal guess is that at least two thirds of the casts in HIPL
could be avoided. Often the internal API is designed in a way that
seems to make casting back and forth a necessity. If this is the
case, then the API is badly designed.

---

I forgot the completely redundant casts: The results of functions returning
foo are cast to .. foo!

As a random example, grep for append_to_slist and witness that it returns
'SList *' but the result is cast to 'SList *'...

firewall/conntrack.c:

    connection->reply.esp_tuples = (SList *) append_to_slist((SList *)

[René Hummen (rene-hummen)]

Diego, can you have a quick look and asses the current situation. I guess casts should be at a level that is manageable.
Please, do _not_ start fixing.

[Diego Biurrun (diego-biurrun)]

The rampant casting is still not under control. Some void* casts are surely still left as are pointless return value casts.

Then there is the deeper issue of data structures that are (mis)designed to only work with casting.