Re: [PATCH 05/17] xen/x86: address violations of MISRA C:2012 Directive 4.10

[Jan Beulich <jbeulich@xxxxxxxx>]

On 13.07.2024 00:09, Stefano Stabellini wrote:
> On Tue, 9 Jul 2024, Jan Beulich wrote:
>> On 09.07.2024 09:38, Alessandro Zucchelli wrote:
>>> On 2024-07-01 16:21, Jan Beulich wrote:
>>>> On 01.07.2024 15:36, Alessandro Zucchelli wrote:
>>>>> --- a/xen/arch/x86/Makefile
>>>>> +++ b/xen/arch/x86/Makefile
>>>>> @@ -260,17 +260,18 @@ $(objtree)/arch/x86/include/asm/asm-macros.h: 
>>>>> $(obj)/asm-macros.i $(src)/Makefil
>>>>>   $(call filechk,asm-macros.h)
>>>>>
>>>>>  define filechk_asm-macros.h
>>>>> + guard=$$(echo ASM_${SRCARCH}_ASM_MACROS_H | tr a-z A-Z);  \
>>>>
>>>> Nit: Hard tab slipped in.
>>>>
>>>>> +    echo '#ifndef $$guard'; \
>>>>> +    echo '#define $$guard'; \
>>>>>      echo '#if 0'; \
>>>>>      echo '.if 0'; \
>>>>>      echo '#endif'; \
>>>>> -    echo '#ifndef __ASM_MACROS_H__'; \
>>>>> -    echo '#define __ASM_MACROS_H__'; \
>>>>>      echo 'asm ( ".include \"$@\"" );'; \
>>>>> -    echo '#endif /* __ASM_MACROS_H__ */'; \
>>>>>      echo '#if 0'; \
>>>>>      echo '.endif'; \
>>>>>      cat $<; \
>>>>> -    echo '#endif'
>>>>> +    echo '#endif'; \
>>>>> +    echo '#endif /* $$guard */'
>>>>>  endef
>>>>>
>>>>>  $(obj)/efi.lds: AFLAGS-y += -DEFI
>>>>> --- a/xen/arch/x86/cpu/cpu.h
>>>>> +++ b/xen/arch/x86/cpu/cpu.h
>>>>> @@ -1,3 +1,6 @@
>>>>> +#ifndef X86_CPU_CPU_H
>>>>> +#define X86_CPU_CPU_H
>>>>
>>>> This, ...
>>>>
>>>>> --- a/xen/arch/x86/x86_64/mmconfig.h
>>>>> +++ b/xen/arch/x86/x86_64/mmconfig.h
>>>>> @@ -5,6 +5,9 @@
>>>>>   * Author: Allen Kay <allen.m.kay@xxxxxxxxx> - adapted from linux
>>>>>   */
>>>>>
>>>>> +#ifndef X86_64_MMCONFIG_H
>>>>> +#define X86_64_MMCONFIG_H
>>>>
>>>> ... this, and ...
>>>>
>>>>> --- a/xen/arch/x86/x86_emulate/private.h
>>>>> +++ b/xen/arch/x86/x86_emulate/private.h
>>>>> @@ -6,6 +6,9 @@
>>>>>   * Copyright (c) 2005-2007 XenSource Inc.
>>>>>   */
>>>>>
>>>>> +#ifndef X86_X86_EMULATE_PRIVATE_H
>>>>> +#define X86_X86_EMULATE_PRIVATE_H
>>>>
>>>> ... this guard can't possibly all follow the same proposed naming 
>>>> scheme
>>>> (wherever the final version of that is being recorded; I don't recall 
>>>> it
>>>> having gone in, and I didn't spot anything earlier in the series); at
>>>> least one must be wrong.
>>>
>>> For x86/x86_64/mmconfig.h has been made an exception as stated in the 
>>> commit
>>> message:
>>> Note that in x86_64/mmconfig.h we slightly deviate from the naming
>>> convention in place: instead of having the inclusion guard as
>>> X86_X86_64_MMCONFIG_H we shortened the directory prefix as X86_64 for
>>> the sake of readability.
>>>
>>> If you do not agree with this exception and you prefer to keep the 
>>> additional
>>> X86_ prefix let me know so as I prepare the patch series V5 I may 
>>> reintroduce it.
>>
>> What I have an issue with is making an exception in one place when quite
>> clearly would as much (or as little) benefit from one. Before there's
>> any further back and forth, I'd like to suggest that you wait with
>> adjustments here until the base scheme has really been agreed upon,
>> including situations where we think we'd like to make exceptions (after
>> all we might decide that there simply shouldn't be exceptions, and that
>> in order to eliminate such redundancy in guard identifiers we'd rather
>> like to rename some of the files).
> 
> I don't think it is a good idea to introduce a naming scheme and
> immediately add exceptions. I would stick to the naming scheme even if
> it doesn't always lead to the best possible name. Also these are header
> guards, they are not critical variables for which clarity and shortness
> are paramount.
> 
> The naming scheme was written in the 0 email and copy/pasted here for
> convenience:
> 
> - private headers -> <dir>_<filename>_H
> - asm-generic headers -> ASM_GENERIC_<filename>_H
> - arch/<architecture>/include/asm/<subdir>/<filename>.h -> 
> ASM_<architecture>_<subdir>_<filename>_H
> - include/generated/<subdir>/<filename>.h-> GENERATED_<subdir>_<filename>_H
> - arch/<architecture>/include/generated/asm/<filename>.h-> 
> <arch>_GENERATED_ASM_<name>_H
> 
> So in this patch I would sticked to the naming scheme and used:
> 
> X86_CPU_CPU_H
> X86_X86_64_MMCONFIG_H
> X86_X86_EMULATE_PRIVATE_H

Uniformly sticking to the scheme is definitely an option. Provided we're
happy with excessively long identifiers (think of, sooner or later, there
going to be style violations wrt line length) and ones having overly
redundant name components. Yet as said elsewhere, first and foremost I
think we need a scheme where collisions cannot occur.

Jan