Re: [PATCH] docs/misra: new rules addition

[Jan Beulich <jbeulich@xxxxxxxx>]

On 12.06.2023 10:58, Roberto Bagnara wrote:
> On 09/06/23 10:46, Jan Beulich wrote:
>> On 08.06.2023 13:02, Roberto Bagnara wrote:
>>> On 07/06/23 23:53, Stefano Stabellini wrote:
>>>> On Wed, 7 Jun 2023, Jan Beulich wrote:
>>>>>> +   * - `Rule 5.6 
>>>>>> <https://gitlab.com/MISRA/MISRA-C/MISRA-C-2012/Example-Suite/-/blob/master/R_05_06.c>`_
>>>>>> +     - Required
>>>>>> +     - A typedef name shall be a unique identifier
>>>>>> +     -
>>>>>
>>>>> Considering that the rule requires uniqueness across the entire code
>>>>> base (and hence precludes e.g. two functions having identically named
>>>>> local typedefs), I'm a little puzzled this was adopted. I for one
>>>>> question that a use like the one mentioned is really at risk of being
>>>>> confusing. Instead I think that the need to change at least one of
>>>>> the names is at risk of making the code less readable then, even if
>>>>> ever so slightly. (All of this said - I don't know if we have any
>>>>> violations of this rule.)
>>>>
>>>> I don't think we have many local typedefs and I think we have only few
>>>> violations if I remember right. I'll let Roberto confirm how many. This
>>>> rule was considered not a difficult rule (some difficult rules were
>>>> found, namely 2.1, 5.5 and 7.4.)
>>>
>>> There currently are 30 violations for ARM64:
>>>
>>> - some involve a typedef module_name (in the call it was said this should
>>>     be renamed module_name_t, which will solve the issue);
>>> - most concern repeated typedefs (UINT64 and similar) which are repeated
>>>     in xen/arch/arm/include/asm/arm64/efibind.h
>>>     and xen/include/acpi/actypes.h
>>> - ret_t and phys_addr_t are also redefined in a couple of places.
>>>
>>> There are 90 violations for X86_64, for the same reasons, plus
>>>
>>> - another set of typedefs for basic types (such as u8);
>>> - repeated typedefs for things like guest_l1e_t in the same header file:
>>>
>>> xen/arch/x86/include/asm/guest_pt.h:60.39-60.49:
>>> for program `xen/.xen-syms.0', the identifier for typedef `guest_l1e_t' is 
>>> reused
>>> xen/arch/x86/include/asm/guest_pt.h:128.22-128.32:
>>> for program `xen/.xen-syms.0', the identifier for typedef `guest_l1e_t' is 
>>> reused
>>>
>>> The indicated lines read as follows:
>>>
>>> typedef struct { guest_intpte_t l1; } guest_l1e_t;
>>> typedef l1_pgentry_t guest_l1e_t;
>>
>> So this is an example where I don't think we can sensibly do away with the
>> re-use of the same typedef name: We use it so we can build the same source
>> files multiple ways, dealing with different paging modes guests may be in.
> 
> Typedefs being used this way can be deviated with tool configuration.
> Here is a list of candidates for that treatment:
> 
> guest_intpte_t
> guest_l1e_t
> guest_l2e_t
> ret_
> 
> I am not sure about the latter.  Please let me know if this is what
> you would prefer and possible additions to/removals from the above list.

Well, if deviating such is possible despite their extended use (in certain
places), then fine. I was afraid that a deviation with such wide a scope
might be hard to justify.

Jan