Re: [PATCH] Makefile: Globally enable fall-through warning

[Guenter Roeck <linux@roeck-us.net>]

Hi Gustavo,

On Sun, Jul 28, 2019 at 11:42:28AM -0500, Gustavo A. R. Silva wrote:
> Hi Guenter,
> 
> On 7/28/19 8:58 AM, Guenter Roeck wrote:
> > On Thu, Jun 06, 2019 at 07:46:17PM -0500, Gustavo A. R. Silva wrote:
> >> Now that all the fall-through warnings have been addressed in the
> >> kernel, enable the fall-through warning globally.
> >>
> > 
> > Not really "all".
> > 
> > powerpc:85xx/sbc8548_defconfig:
> > 
> > arch/powerpc/kernel/align.c: In function ‘emulate_spe’:
> > arch/powerpc/kernel/align.c:178:8: error: this statement may fall through
> > 
> > Plus many more similar errors in the same file.
> > 
> > All sh builds:
> > 
> > arch/sh/kernel/disassemble.c: In function 'print_sh_insn':
> > arch/sh/kernel/disassemble.c:478:8: error: this statement may fall through
> > 
> > Again, this is seen in several places.
> > 
> > mips:cavium_octeon_defconfig:
> > 
> > arch/mips/cavium-octeon/octeon-usb.c: In function 'dwc3_octeon_clocks_start':
> > include/linux/device.h:1499:2: error: this statement may fall through
> > 
> > None of those are from recent changes. And this is just from my small
> > subset of test builds.
> > 
> 
> Thank you for letting me know about this. I don't have access to build
> infrastructure like yours.
> 

I am always happy to run test builds on my infrastructure.

> My build infrastructure is similar to that of Linus.
> 
> But if you send me all of those I can create a patch and send it back
> to you to make sure what you see is addressed. If we can coordinate for
> this it'd be great for everybody. :)
> 

Just have a look at the output of https://kerneltests.org/builders/,
in the 'master' and/or 'next' column. There are many additional warnings
in 'next'. Only downside is that you won't see the warnings unless there
are also build errors, but -next tends to have lots of those.

Just wondering ... wouldn't it be possible to run a coccinelle script
to identify those problems automatically, without depending on compile
warnings ? Or smatch/sparse, maybe ?

Thanks,
Guenter