Re: [tip:x86/build] x86/build: Remove -Wno-sign-compare

[Ingo Molnar <mingo@kernel.org>]

* Valdis.Kletnieks@vt.edu <Valdis.Kletnieks@vt.edu> wrote:

> On Tue, 12 May 2015 10:44:15 +0200, Ingo Molnar said:
> 
> > Btw., just some feedback, 'random' kernel configs still generate a ton
> > of warnings:
> >
> >  randconf: #   9, ed602bbb, Tue_May_12_09_07_25_CEST_2015:  39 kernels/hour,
> [ bzImage...    81 secs, modules...    21 secs, done ] (warns:   6)
> >  randconf: #  10, ed602bbb, Tue_May_12_09_10_10_CEST_2015:  36 kernels/hour,
> [ bzImage...    43 secs, modules...    20 secs, done ] (warns:  12)
> >  randconf: #  11, ed602bbb, Tue_May_12_09_11_52_CEST_2015:  36 kernels/hour,
> [ bzImage...    71 secs, modules...     0 secs, done ] (warns:   5)
> >  randconf: #  12, ed602bbb, Tue_May_12_09_12_56_CEST_2015:  37 kernels/hour,
> [ bzImage...    64 secs, modules...    28 secs, done ] (warns:  16)
> >  randconf: #  13, ed602bbb, Tue_May_12_09_14_07_CEST_2015:  38 kernels/hour,
> [ bzImage...   106 secs, modules...     0 secs, done ] (warns: 157)
> >  randconf: #  14, ed602bbb, Tue_May_12_09_15_40_CEST_2015:  38 kernels/hour,
> [ bzImage...   100 secs, modules...     0 secs, done ] (warns:  11)
> >  randconf: #  15, ed602bbb, Tue_May_12_09_17_27_CEST_2015:  37 kernels/hour,
> [ bzImage...    65 secs, modules...    28 secs, done ] (warns:  26)
> >  randconf: #  16, ed602bbb, Tue_May_12_09_19_08_CEST_2015:  37 kernels/hour,
> [ bzImage...    70 secs, modules...     0 secs, done ] (warns: 228)
> 
> > Before I pushed out this -Wno-sign-compare change I made sure there
> > are no extra warnings generated on the 8 key configs I monitor
> > explicitly: [def|allno|allyes|allmod]config[64|32].
> 
> It makes my config totally explode on next-20150512
> 
> % gcc --version
> gcc (GCC) 5.1.1 20150422 (Red Hat 5.1.1-1)
> 
> % grep "warning: comparison between signed and unsigned integer expressions" build.default | wc -l
> 64148
> 
> A *lot*of them appear to be exploding inside likely().  Here's all the
> exploding for *one file*...

That's a prerelease of GCC, right? So I think GCC gets constant 
propagation from various builtins wrong or so.

But ... the output looks horrible, and for years we didn't have the 
warnings, still after reintroducing it we didn't get any new warnings 
about truly bogus code.

So it does not seem to have much utility, and seems to be horribly 
broken on fresh GCC.

I'll remove the commit.

Thanks,

	Ingo