Re: [PATCH 1/2] automation/eclair_analysis: deviate MISRA C Rule 21.2

[Jan Beulich <jbeulich@xxxxxxxx>]

On 21.06.2024 03:02, Stefano Stabellini wrote:
> On Thu, 20 Jun 2024, Jan Beulich wrote:
>> On 19.06.2024 19:09, Alessandro Zucchelli wrote:
>>> Rule 21.2 reports identifiers reserved for the C and POSIX standard
>>> libraries: all xen's translation units are compiled with option
>>> -nostdinc, this guarantees that these libraries are not used, therefore
>>> a justification is provided for allowing uses of such identifiers in
>>> the project.
>>> Builtins starting with "__builtin_" still remain available.
>>>
>>> No functional change.
>>>
>>> Signed-off-by: Alessandro Zucchelli <alessandro.zucchelli@xxxxxxxxxxx>
>>> ---
>>>  automation/eclair_analysis/ECLAIR/deviations.ecl | 11 +++++++++++
>>>  1 file changed, 11 insertions(+)
>>>
>>> diff --git a/automation/eclair_analysis/ECLAIR/deviations.ecl 
>>> b/automation/eclair_analysis/ECLAIR/deviations.ecl
>>> index 447c1e6661..9fa9a7f01c 100644
>>> --- a/automation/eclair_analysis/ECLAIR/deviations.ecl
>>> +++ b/automation/eclair_analysis/ECLAIR/deviations.ecl
>>> @@ -487,6 +487,17 @@ leads to a violation of the Rule are deviated."
>>>  # Series 21.
>>>  #
>>>  
>>> +-doc_begin="Rules 21.1 and 21.2 report identifiers reserved for the C and 
>>> POSIX
>>> +standard libraries: if these libraries are not used there is no reason to 
>>> avoid such
>>> +identifiers. All xen's translation units are compiled with option 
>>> -nostdinc,
>>> +this guarantees that these libraries are not used. Some compilers could 
>>> perform
>>> +optimization using built-in functions: this risk is partially addressed by
>>> +using the compilation option -fno-builtin. Builtins starting with 
>>> \"__builtin_\"
>>> +still remain available."
>>
>> While the sub-section "Reserved Identifiers" is part of Section 7,
>> "Library", close coordination is needed between the library and the
>> compiler, which only together form an "implementation". Therefore any
>> use of identifiers beginning with two underscores or beginning with an
>> underscore and an upper case letter is at risk of colliding not only
>> with a particular library implementation (which we don't use), but
>> also of such with a particular compiler implementation (which we cannot
>> avoid to use). How can we permit use of such potentially problematic
>> identifiers?
> 
> Alternative question: is there a way we can check if there is clash of
> some sort between a compiler implementation of something and a MACRO or
> identifier we have in Xen? An error or a warning from the compiler for
> instance? That could be an easy way to prove we are safe.

Well. I think it is the default for the compiler to warn when re-#define-
ing a previously #define-d (by the compiler or by us) symbol, so on that
side we ought to be safe at any given point in time, yet we're still
latently unsafe (as to compilers introducing new pre-defines). For
built-in declarations, though, there's nothing I'm aware of that would
indicate collisions.

> Also, can we use the fact that the compiler we use is the same compiler
> used to compile Linux, and Linux makes extensive use of identifiers and
> macros starting with underscores as one of the reason for being safe
> from clashes?

I think we could, but I don't think we should.

Jan