[PATCH] d3dx9: Don't check for negative enum value.
jacek at codeweavers.com
Tue Mar 24 09:50:31 CDT 2020
On 24.03.2020 15:39, Matteo Bruni wrote:
> On Tue, Mar 24, 2020 at 3:29 PM Jacek Caban <jacek at codeweavers.com> wrote:
>> On 24.03.2020 14:59, Henri Verbeet wrote:
>>> On Tue, 24 Mar 2020 at 17:45, Matteo Bruni <matteo.mystral at gmail.com> wrote:
>>>> I assume replacing "op < 0" with an explicit "op < SCT_VSFLOAT"
>>>> doesn't avoid the warning. Does replacing the if with an assert do the
>>>> trick, by any chance?
>>> It's perhaps a little subtle, but if you compare "op" against
>>> "ARRAY_SIZE(const_tbl)" (which is what you really care about here
>>> anyway, right?) instead of SCT_PSINT, you can drop the check against
>>> 0, regardless of whether the enum ends up being a signed or unsigned
>> Strictly speaking, if enum would be signed the check wouldn't catch some
>> invalid values, but I don't think that's worrying about. If we want it
>> fully strict, we could change argument type from enum to unsigned int.
> I think sizeof() (and thus ARRAY_SIZE()) always returns an unsigned
> integer, op is promoted to unsigned if necessary and the comparison is
> guaranteed to be unsigned.
Sure, but theoretically if underlying type of enum is signed char
(actually, signed part is not important), then it will be truncated
before being promoted.
More information about the wine-devel