Re: [PATCH v8 07/10] lib: add basic KUnit test for lib/math

[Andy Shevchenko]

On Mon, May 20, 2024 at 07:51:24PM +0530, Devarsh Thakkar wrote:
> On 20/05/24 17:52, Andy Shevchenko wrote:
> > On Mon, May 20, 2024 at 05:11:18PM +0530, Devarsh Thakkar wrote:
> >> On 18/05/24 01:44, Andy Shevchenko wrote:
> >>> On Fri, May 17, 2024 at 11:06:07PM +0530, Devarsh Thakkar wrote:

[..]

> > Yes, and one should follow IWYU principle and not cargo cult or whatever
> > arbitrary lists.
> 
> Agreed.

> >>>> +#include <linux/lcm.h>
> >>>
> >>> + math.h // obviously
> >>> + module.h
> >>>
> >>>> +#include <linux/reciprocal_div.h>
> >>>
> >>> + types.h
> >>
> >> All the above headers are already included as part of kernel.h
> > 
> > Yes, that's why you should not use "proxy" headers.
> > Have you read the top comment in the kernel.h?
> 
> Yes, it says it is not recommended to include this inside another header file.
> Although here we are adding it inside c file, but I can still try avoid it and
> include only the required headers instead of kernel.h as you recommended.

Right, but the first sentence there is
"This header has combined a lot of unrelated to each other stuff."

Can you explain how you use in your code all that unrelated stuff?
For example, how do you use *trace_*() calls? Or maybe might_*() calls?
or anything else that is directly provided by kernel.h?

Besides IWYU principle above, it's good to have a justification for each
inclusion the C file has. I believe there is no a such in _this_ case.

-- 
With Best Regards,
Andy Shevchenko