Re: [PATCH 1/4] x86/domain: address Misra rule 11.1 violation in reset_stack_and_call_ind()

[Jan Beulich <jbeulich@xxxxxxxx>]

On 12.09.2026 17:09, Nicola Vetrini wrote:
> On 2026-09-03 13:43, Jan Beulich wrote:
>> Eclair dislikes both sides of the comparison to differ in noreturn
>> attributes, thus deeming this a violation of "Conversions shall not be
>> performed between a pointer to a function and any other type". Since gcc
>> doesn't permit use of "noreturn" types used in a typecast, resort to
>> typeof().
>>
>> No functional change intended.
>>
>> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> 
> Reviewed-by: Nicola Vetrini <nicola.vetrini@xxxxxxxxxxx>

Thanks.

>> ---
>> Really Eclair's diagnosis is misleading here, and not only because of the
>> missing "noreturn" there, making both sides be identical: There are no
>> implicit conversions performed on this kind of pointer operands of an
>> equality expression. A diagnosis of "comparison of pointers to different
>> types", along the lines of what compilers use, would be more to the point.
> 
> The point here is that the types are not identical, if you consider noreturn 
> as part of the function type (as clang does) or not (as GCC does). This is 
> the reason why we err on the side of caution and diagnose this, and it's not 
> likely to change in the near future. Perhaps we could make the violation 
> message more explicit about the motivation of the mismatch.

Yes, they are different by "noreturn", but that's completely invisible from
the report. That's the whole point of the remark: By its wording, the report
at the same time says there is a violation just to use what's quoted to
indicate there's none.

Jan