Re: [PATCH 08/46] kernel: MOve prototype declaration to header file include/linux/perf_event.h

[Peter Zijlstra]

On Thu, Feb 27, 2014 at 02:55:45PM -0800, josh@xxxxxxxxxxxxxxxx wrote:
> On Thu, Feb 27, 2014 at 08:23:35PM +0100, Peter Zijlstra wrote:
> > On Thu, Feb 27, 2014 at 07:51:50AM -0800, Josh Triplett wrote:
> > > On Thu, Feb 27, 2014 at 12:54:14PM +0100, Peter Zijlstra wrote:
> > > > On Thu, Feb 27, 2014 at 05:02:48PM +0530, Rashika Kheria wrote:
> > > > > Add prototype declaration of function to header file
> > > > > include/linux/perf_event.h because it is used by more than one file.
> > > > > 
> > > > > This eliminates the following warning in kernel/events/core.c:
> > > > > kernel/events/core.c:3743:13: warning: no previous prototype for âarch_perf_update_userpageâ [-Wmissing-prototypes]
> > > > 
> > > > # git grep arch_perf_update_userpage
> > > > arch/x86/kernel/cpu/perf_event.c:void arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now)
> > > > kernel/events/core.c:void __weak arch_perf_update_userpage(struct perf_event_mmap_page *userpg, u64 now)
> > > > kernel/events/core.c:   arch_perf_update_userpage(userpg, now);
> > > > 
> > > > 
> > > > There's two definitions; one weak, and one usage site.
> > > > 
> > > > What gives?
> > > 
> > > There's no prototype for the function anywhere, so -Wmissing-prototypes
> > > rightfully complains.  Adding the prototype to a header included in both
> > > source files ensures that the function signatures must match, and
> > > eliminates the warning.
> > 
> > Definitions don't require prior declarations. Only usage without prior
> > definitions require them.
> > 
> > I still don't see a problem.
> 
> There's value in -Wmissing-prototypes; it's equivalent to the Sparse
> check that ensures functions are marked as static when they're not used
> outside the file they're defined in (though unlike the Sparse check it
> only applies to functions, not data).  The goal isn't "make the warning
> go away"; the goal of passing -Wmissing-prototypes is:
> 
> - Get rid of unused functions (many of which have been caught by this
>   effort).
> - Mark functions as static where possible (which enables the above and
>   also improves code generation).
> - Include headers that declare functions in the source files that define
>   them, not just in the source files that use them.  That keeps the
>   declaration and definition in sync (which has caught some real bugs as
>   part of this effort).
> - When no such header exists for a function used across multiple files,
>   add the prototype to an appropriate header and use that.  Again, this
>   has caught real bugs as part of this effort, when the definition and
>   use were out of sync.  (For instance, disagreeing in return type.)

Well then say that already;  the reason I reacted so strongly is because
I'm busy, and I don't have time for */46 patches that look like robot
output.

The changelog reads like its just making warns go away; the patches look
automated; including the typical non-thinking robot fails; _AND_ its
got a review tag from someone who should bloody well know better than to
send crap patches like this.

This all is a huge waste of time from my side; and you're doing your
intern a disservice by training him to be a thoughtless robot.


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/