Re: [XEN PATCH] xen/compiler: deviate the inline macro for MISRA C Rule 20.4

[Jan Beulich <jbeulich@xxxxxxxx>]

On 11.03.2024 16:48, Nicola Vetrini wrote:
> On 2024-03-11 08:32, Jan Beulich wrote:
>> On 08.03.2024 09:10, Nicola Vetrini wrote:
>>> --- a/docs/misra/safe.json
>>> +++ b/docs/misra/safe.json
>>> @@ -28,6 +28,14 @@
>>>          },
>>>          {
>>>              "id": "SAF-3-safe",
>>> +            "analyser": {
>>> +                "eclair": "MC3R1.R20.4"
>>> +            },
>>> +            "name": "MC3R1.R20.4: allow the augmentation of the 
>>> inline keyword in some build configurations",
>>> +            "text": "The definition of this macro named inline allows 
>>> further checking in some build configurations that certain symbols end 
>>> up in the right sections."
>>> +        },
>>
>> With this wording the ID isn't going to be re-usable anywhere else. 
>> Even
>> if the exact same reasoning would apply.
>>
> 
> What about
> 
> "name": "MC3R1.R20.4: allow the definition of a macro with the same name 
> as a keyword in some special cases"
> 
> and
> 
> "text": "The definition of a macro with the same name as a keyword can 
> be useful in certain configurations to improve the guarantees that can 
> be provided by Xen. See docs/misra/deviations.rst for a precise 
> rationale for all such cases.
> 
> and then..
> 
>>> +        {
>>> +            "id": "SAF-4-safe",
>>>              "analyser": {},
>>>              "name": "Sentinel",
>>>              "text": "Next ID to be used"
>>> diff --git a/xen/include/xen/compiler.h b/xen/include/xen/compiler.h
>>> index 16d554f2a593..e3d9f9fb8e4b 100644
>>> --- a/xen/include/xen/compiler.h
>>> +++ b/xen/include/xen/compiler.h
>>> @@ -82,6 +82,7 @@
>>>   * inline functions not expanded inline get placed in .init.text.
>>>   */
>>>  #include <xen/init.h>
>>> +/* SAF-3-safe MISRA C Rule 20.4: define the inline macro to perform 
>>> checks */
>>>  #define inline inline __init
>>>  #endif
>>
>> I don't think the definition is "to perform checks"; it's rather to 
>> make
>> sure checking elsewhere wouldn't (seemingly) randomly fail. 'Override
>> "inline" for section contents checks to pass when the compiler chooses
>> not to inline a particular function' perhaps? Albeit that's getting
>> long-ish, I fear.
> 
> put this message in deviations.rst
> 
> is this proposal more appealing?

I think so, yes.

Jan