Re: [PATCH] x86/fpu: fix boolreturn.cocci warnings

From: Ingo Molnar <mingo@kernel.org>
To: Thomas Gleixner <tglx@linutronix.de>
Cc: kbuild test robot <fengguang.wu@intel.com>,
	Rik van Riel <riel@redhat.com>,
	kbuild-all@01.org, LKML <linux-kernel@vger.kernel.org>,
	tipbuild@zytor.com, "H. Peter Anvin" <hpa@zytor.com>,
	x86@kernel.org, Dave Hansen <dave.hansen@linux.intel.com>,
	Yu-cheng Yu <yu-cheng.yu@intel.com>,
	Fenghua Yu <fenghua.yu@intel.com>, Borislav Petkov <bp@suse.de>,
	Peter Zijlstra <peterz@infradead.org>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Andrew Morton <akpm@linux-foundation.org>
Subject: Re: [PATCH] x86/fpu: fix boolreturn.cocci warnings
Date: Tue, 7 Mar 2017 10:01:44 +0100	[thread overview]
Message-ID: <20170307090144.GB6946@gmail.com> (raw)
In-Reply-To: <alpine.DEB.2.20.1703070924220.3584@nanos>

(Linus and Andrew Cc:-ed)

* Thomas Gleixner <tglx@linutronix.de> wrote:

> On Tue, 7 Mar 2017, Ingo Molnar wrote:
> > 
> > * kbuild test robot <fengguang.wu@intel.com> wrote:
> > 
> > > arch/x86/kernel/fpu/xstate.c:931:9-10: WARNING: return of 0/1 in function 'xfeatures_mxcsr_quirk' with return type bool
> > > 
> > >  Return statements in functions returning bool should use
> > >  true/false instead of 1/0.
> > 
> > Note that this is a totally bogus warning. I personally find a 0/1 return more 
> > readable than a textual 'true/false', even if bools are used, and nowhere does the 
> > kernel mandate the use of 0/1.
> 
> I disagree.
> 
> The fact that booleans have been brought retroactively into the C-Standard
> does and for compability reasons C still follows the approach "Boolean
> values are just integers" does not make it any better.
> 
> We had stupid bugs, where people returned -EINVAL from a boolean function
> and introduced silly and hard to understand bugs.

But this function is not using -EINVAL, it's using 0 and 1 which is both correct 
and unambiguous!

I mean, if the Cocci script warned about -EINVAL then it would have found a clear 
bug. Now it's warning about the use of 0/1 literals with bool types which is 
perfectly legal, readable, clear C code!

> The canonical values assigned to booleans are 'true' and 'false' and not 
> whatever people prefer. Can we please be consistent on that?

I think that's backwards, because 1/0 is just as canonical for true/false, and to 
me personally it's in fact easier to read as well.

I would really like higher level buy-in for that principle (I've Cc:-ed Linus and 
Andrew), and if indeed the consensus is that '0/1' cannot be used with 'bool' then 
I'll remove all uses of 'bool' from my patches and from code I care about and use 
'int' instead. Please update Documentation/CodingStyle accordingly as well.

To me a lexical 'true/false' instead of '1/0' is a step backwards in readability 
in many cases - using the slightly wider 'int' type is the lesser evil.

Thanks,

	Ingo