Re: [PATCH *-next 00/18] Remove weird and needless 'return' for void APIs

[Peter Zijlstra <peterz@infradead.org>]

On Thu, Feb 27, 2025 at 08:48:19PM +0800, Zijun Hu wrote:
> On 2025/2/21 21:02, Zijun Hu wrote:
> > void api_func_a(...);
> > 
> > static inline void api_func_b(...)
> > {
> > 	return api_func_a(...);
> > }
> 
> The Usage : Return void function in void function
> 
> IMO, perhaps, the usage is not good since:
> 
> A) STD C does not like the usage, and i find GCC has no description
> about the usage.
>    C11/C17: 6.8.6.4 The return statement
>    "A return statement with an expression shall not appear in a
> function whose return type is void"

We really don't use STD C, the kernel is littered with extensions.

> B) According to discussion, the usage have function that return type
>    of the callee api_func_a() is monitored. but this function has below
> shortcoming as well:
> 
> the monitor is not needed if the caller api_func_b() is in the same
> module with the callee api_func_a(), otherwise, provided the callee is
> a API and provided by author of other module. the author needs to clean
> up lot of usages of the API if he/she changes the API's return type from
> void to any other type, so it is not nice to API provider.
> 
> C) perhaps, most ordinary developers don't known the function mentioned
>    by B), and also feel strange for the usage

It is quite common to do kernel wide updates using scripts / cocinelle.

If you have a specialization that wraps a function to fill out a default
value, then you want the return types to keep matching.

Ex.

return_type foo(type1 a1, type2 a2);

return_type my_foo(type1 a1)
{
	return foo(a1, value);
}

is a normal thing to do. The whole STD C cannot return void bollocks
breaks that when return_type := void, so in that regards I would call
this a STD C defect.