Re: [PATCH v4 08/11] xen/compiler: import 'fallthrough' keyword from linux

[Jan Beulich <jbeulich@xxxxxxxx>]

On 13.01.2021 00:30, Stefano Stabellini wrote:
> On Tue, 12 Jan 2021, Jan Beulich wrote:
>> On 08.01.2021 15:46, Rahul Singh wrote:
>>> -Wimplicit-fallthrough warns when a switch case falls through. Warning
>>> can be suppress by either adding a /* fallthrough */ comment, or by
>>> using a null statement: __attribute__ ((fallthrough))
>>
>> Why is the comment variant (which we use in many places already,
>> albeit with varying wording) not the route of choice?
> 
> See previous discussion:
> 
> https://marc.info/?l=xen-devel&m=160707274517270
> https://marc.info/?l=xen-devel&m=160733742810605
> https://marc.info/?l=xen-devel&m=160733852011023
> 
> We thought it would be best to introduce "fallthrough" and only resort
> to comments as a plan B. The usage of the keyword should allow GCC to do
> better checks.

Hmm, this earlier discussion was on an Arm-specific thread, and I
have to admit I can't see arguments there pro and/or con either
of the two alternatives.

>>> Define the pseudo keyword 'fallthrough' for the ability to convert the
>>> various case block /* fallthrough */ style comments to null statement
>>> "__attribute__((__fallthrough__))"
>>>
>>> In C mode, GCC supports the __fallthrough__ attribute since 7.1,
>>> the same time the warning and the comment parsing were introduced.
>>>
>>> fallthrough devolves to an empty "do {} while (0)" if the compiler
>>> version (any version less than gcc 7) does not support the attribute.
>>
>> What about Coverity? It would be nice if we wouldn't need to add
>> two separate constructs everywhere to make both compiler and static
>> code checker happy.
> 
> I don't think I fully understand your reply here: Coverity doesn't come
> into the picture. Given that GCC provides a special keyword to implement
> fallthrough, it makes sense to use it when available. When it is not
> available (e.g. clang or older GCC) we need to have an alternative to
> suppress the compiler warnings. Hence the need for this check:
> 
>   #if (!defined(__clang__) && (__GNUC__ >= 7))

I'm not sure how this interacts with Coverity. My point bringing up
that one is that whatever gets done here should _also_ result in
Coverity recognizing the fall-through as intentional, or else we'll
end up with many unwanted reports of new issues once the pseudo-
keyword gets made use of. The comment model is what we currently
use to "silence" Coverity; I'd like it to be clear up front that
any new alternative to be used is also going to "satisfy" it.

Jan