Re: [PATCH] x86/DMI: adjustments to comply with Misra C:2012 Rule 9.3

[Nicola Vetrini <nicola.vetrini@xxxxxxxxxxx>]

On 2023-11-30 08:55, Jan Beulich wrote:
> The rule demands that all array elements be initialized (or dedicated
> initializers be used). Introduce a small set of macros to allow doing 
> so
> without unduly affecting use sites (in particular in terms of how many
> elements .matches[] actually has; right now there's no use of
> DMI_MATCH4(), so we could even consider reducing the array size to 3).
>
> Note that DMI_MATCH() needs adjustment because of the comma included in
> its expansion, which - due to being unparenthesized - would otherwise
> cause macro arguments in the "further replacement" step to be wrong.
>
> Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> ---
> Of course a question is how many of these DMI table entries are in fact
> no longer applicable (e.g. because of naming 32-bit-only systems).
> Subsequently the table in dmi_scan.c itself may want cleaning up as
> well, yet I guess the question of stale entries is even more relevant
> there.
>
> An alternative to using the compound literal approach might be to go
> along the lines of
>
> #define DMI_MATCH4(m1, m2, m3, m4) .matches = { [0] = m1, [1] = m2, [2] 
> = m3, [3] = m4 }
> I've chosen the other approach mainly because of being slightly 
> shorter.
From a MISRA perspective this resolves all but one violation: the 
initialization of static array
ns16550_com[2] = { { 0 } };

in drivers/char/ns16550.c. This is a case where the explicit initializer 
looks unnecessary.
--
Nicola Vetrini, BSc
Software Engineer, BUGSENG srl (https://bugseng.com)