Re: Basic C question

[Rene Herman <rene.herman@xxxxxxxxxxxx>]

On 03-08-08 09:58, Manish Katiyar wrote:

> ok... I know i am going to be embarrassed but I am confused by the 
> below function.

I'd not be embarrased...

> The last parameter in do_mpage_readpage() is of type get_block_t, but
> when passed to block_read_full_page() it gets passed as get_block_t *

The exact nature of get_block_t is important here. Variables of type 
get_block_t are function pointers -- have function type:

typedef int (get_block_t)(...);

> static struct bio * do_mpage_readpage(............., get_block_t get_block) {
>        ................
>        ................
>        	if (page_has_buffers(page))
> 		goto confused;
>         .................
> confused:
>       	if (!PageUptodate(page))
> 	        block_read_full_page(page, get_block);
>        .........
> }
>
> int block_read_full_page(struct page *page, get_block_t *get_block) {
>         .........
>        ..........
> }
>
>
> How does this work ? :-( ....... Isn't gcc supposed to catch this and
> flag warning unless I am missing something very obvious.

In C, except as an operand of sizeof or unary &, an expression having 
function type is automatically converted from function type to pointer 
to function.

Or put differently: except for sizeof and unary &, "function_designator" 
and "&function_designator" are interchangeable as expressions. You can 
therefore write the call as:

	block_read_full_page(page, &get_block);

without changing anything, after which I suppose you'd be fine with the 
declarations.

Now, granted, this specific example is still not nice. I haven't checked 
other uses but it appears that block_read_full_page() should've been 
declared not with a get_block_t *, but simply with a get_block_t. If you 
can convince yourself of such, you could submit it...

Rene.


--
To unsubscribe from this list: send an email with
"unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx
Please read the FAQ at http://kernelnewbies.org/FAQ