Re: [GIT PULL] Driver core patches for 5.1-rc1

[Joe Perches <joe@perches.com>]

On Wed, 2019-03-06 at 15:47 -0800, Linus Torvalds wrote:
> On Wed, Mar 6, 2019 at 2:33 AM Greg KH <gregkh@linuxfoundation.org> wrote:
> > Joe Perches (1):
> >       device.h: Add __cold to dev_<level> logging functions
> 
> This is very funky, but that commit generates a new warning in a
> totally unrelated area:

Very very funky.

Are you sure it's the __cold marking of an
entirely unrelated function that isn't
even used in the code with the new warning?

btw; what compiler version?

>  drivers/iio/adc/qcom-pm8xxx-xoadc.c: In function ‘pm8xxx_xoadc_probe’:
>   drivers/iio/adc/qcom-pm8xxx-xoadc.c:633:8: warning: ‘ch’ may be used
> uninitialized in this function [-Wmaybe-uninitialized]
>     ret = pm8xxx_read_channel_rsv(adc, ch, AMUX_RSV4,
>           ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>              &read_nomux_rsv4, true);
>              ~~~~~~~~~~~~~~~~~~~~~~~
>   drivers/iio/adc/qcom-pm8xxx-xoadc.c:426:27: note: ‘ch’ was declared here
>     struct pm8xxx_chan_info *ch;
>                              ^~
> 
> and it all looks entirely insane if you look at that line 633 where
> the ostensibly uninitialized variable is (it clearly _is_ initialized
> there), but if you then look at that line 426 you notice that it
> actually makes some kind of sense. The value comes from another
> function that was apparently inlined, and that other function does not
> "obviously" initialize it.
> 
> I wonder why this wasn't seen in linux-next? Yes, the connection is
> odd, and maybe it's very compiler version dependent, but I do hope
> people react to new warnings. The kernel is entirely warning-free for
> me for an x86-64 allmodconfig build, and I want to keep it that way.
> 
> And _because_ I want to keep it that way (one of the things I do
> during the merge window is look for oddities coming in during pulls,
> and new warnings is a big deal for me), I applied the attached patch.
> Just FYI.
> 
>                    Linus