Hi Alex,

Alejandro Colomar <alx@kernel.org> writes:

> No, I don't like that.  The type on its own line is fine IMO, but in the
> same line as the attribute looks weird.
>
> To me, OK:
>
> 	[[attr]] int foo(void);
>
> 	[[attr]]
> 	int foo(void);
>
> 	[[attr]]
> 	int
> 	foo(void); // I use this style when writing function definitions
>
> To me, not OK:
>
> 	[[attr]] int
> 	foo(void);

Okay, good to know the conventions. Thanks.

> Thanks!  It might be good to CC libc-alpha@ and linux-api@ in such a
> patch.  Or maybe it's overkill.

I will avoid it. I don't expect the syscalls to be removed as many old
programs still use them. And they work fine in most programs where
nanosecond resolution does not matter.

Also, currently glibc does not mark them as
[[deprecated]]/__attribute__ ((__deprecated__)) and I hope that it stays
that way for the foreseeable future. Apple, for example, marks functions
that are fine and used safely with deprecated and it causes many
annoying warnings when building programs. Specifically, any file that
uses sprintf or vfork will cause a warning on MacOS.

Anyways, submitted a v2 for patch 1/3 [1]. The other patches don't
depend on it. Just other things I have noticed.

Collin

[1] https://lore.kernel.org/linux-man/2d37d5a9251af3c1d25cf8e73e3585a9955d5772.1750373011.git.collin.funk1@gmail.com/T/#u