Re: Implicit type conversion check

["Manuel LÃpez-IbÃÃez" <lopezibanez at gmail dot com>]

On 30/11/06, Pierre Chatelier <pierre.chatelier@club-internet.fr> wrote:
> Hello,
>
> I am sorry if it is some usual request, but I could not find recent
> information about that, neither in the mailing list nor in the FAQ.
> Is there any plan to enforce warnings about implicit type conversion ?

I am happy to inform you that there is [1].

Part of this it's already on mainline, you could download it [2]  and
install it [3] in order to test it. Comments are welcome.

[1] http://gcc.gnu.org/wiki/NewWconversion
[2] http://gcc.gnu.org/svn.html
[3] http://gcc.gnu.org/install

> the -W option implies some checks about _comparison_ between signed
> and unsigned, but nothing seems available for _affectation_.

affectation ?

> In the snippet below, only line 9 can trigger a warning.
>
>        1 void function(float f, double d, int i, unsigned int u)
>        2 {
>        3   f = d; /* potential loss of information */
warned by Wconversion
>        4   i = f; /* ~                             */
likewise
>        5   i = d; /* ~                             */
likewise
>        6   u = i; /* may overflow                  */
likewise
>        7   i = u; /* may overflow                  */
likewise.
>        8
>        9   i = (i<u) ? u : i; /* this comparison *does* trigger a
> warning */
if you use Wconversion you will get another warning here, since the
value of the second u is converted to signed int so you may end up
with a negative number in i even if u and i are positive numbers.

Cheers,

Manuel.