[cfe-commits] [Patch] Remove checking signs of operands of conditional operator from -Wsign-compare

[Chandler Carruth]

On Thu, Jul 14, 2011 at 4:01 PM, Richard Trieu <rtrieu at google.com> wrote:

> On Thu, Jul 14, 2011 at 3:53 PM, Chandler Carruth <chandlerc at google.com>wrote:
>
>> On Thu, Jul 14, 2011 at 3:35 PM, Richard Trieu <rtrieu at google.com> wrote:
>>
>>> Currently, -Wsign-compare will warn if conditional operands differ in
>>> signedness. This patch will move this warning to -Wmixed-sign-conditional
>>
>>
>> Why not -Wsign-conversion ? Just wondering why we need a separate flag
>> here.
>>
> -Wsign-conversion has a separate warning that will override this one.
>  Essentially, the conditional statement can have only have one type, so if
> the operands are of different signs, one will be forced to convert, which
> throws the warning.  Here's an example:
>
> void f(unsigned int u, int i) {
>   (void) (true ? u : i);
> }
>
> clang -Wsign-conversion
> warning: operand of ? changes signedness: 'int' to 'unsigned int'
>       [-Wsign-conversion]
>   (void) (true ? u : i);
>                ~     ^
>

Ah, cool


> clang -Wmixed-sign-conditional
> warning: operands of ? are integers of different signs:
>       'unsigned int' and 'int' [-Wmixed-sign-conditional]
>   (void) (true ? u : i);
>                ^ ~   ~
>

So what's the motivation for being able to get *just* these warnings out?
The only bug I'm aware of was specific to -Wsign-compare warning on this
construct in a compile that didn't use -Wsign-conversion, and so expected
the code to compile without warnings. If -Wsign-conversion already catches
this, it would seem to me that that's sufficient?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20110714/fad25c6d/attachment.html>