[Guy Harris <gharris@xxxxxxxxx>]

Francisco Alcoba (TS/EEM) wrote:

> So, if I have
>
> 	address tmp_src, tmp_dst;
> 	guint32 ipv4_address;
>
> and do
>       g_free(tmp_src.data);
> or
> 	g_memmove(tmp_src.data,&ipv4_address,4);
>
> I get both warnings in a row. I do not have them if I do
>
>  	g_free((void *)tmp_src.data)
>
> but I'm not sure this is OK -meaning, it will do what I need without breaking anything
> else-.

That's the best you can do when freeing data, as far as I know; it's 
what we do elsewhere.

The underlying problem is that there is, as far as I know, no good way 
in C to indicate that some piece of data is immutable, i.e. once it's 
been initialized its value can't be changed, but can be freed, so I 
think all you can do is throw in a cast to indicate that it's OK to free 
the data.

For initializing the data, I'd do

	guint32 *src_data;

	src_data = g_malloc(sizeof *src_data);
	*src_data = ipv4_address;
	tmp_src.data = &src_data;

> I've found a reference to a similar situation, that claims this to be some sort 
> of VC6-specific behaviour,

The claim isn't true - the idea of warning about a const pointer being 
passed to a routine where the matching argument isn't VC6-specific, only 
the text of the message is:

	$ cat foo.c
	#include <stdlib.h>

	void
	freeit(const char *p)
	{
		free(p);
	}
	$ gcc -c -O2 -Wall -W foo.c
	foo.c: In function `freeit':
	foo.c:6: warning: passing arg 1 of `free' discards qualifiers from 
pointer target type