[l2-mtd:master 25/29] drivers/mtd/nand/nand_timings.c:45:29: sparse: constant 250000000000 is so big it is long

[Brian Norris]

+ linux-mtd, lkml, linux-sparse

On Tue, Jul 22, 2014 at 10:28:13AM +0200, Boris BREZILLON wrote:
> On Tue, 22 Jul 2014 11:41:08 +0800 kbuild test robot <fengguang.wu at intel.com> wrote:
> > tree:   git://git.infradead.org/users/dedekind/l2-mtd.git master
> > head:   e4e069347030ab0fea7304d06704fb3b64e5fe40
> > commit: 974647ea8a13021a91d558df61d598bcabf73439 [25/29] mtd: nand: add ONFI timing mode to nand_timings converter
> > reproduce: make C=1 CF=-D__CHECK_ENDIAN__
> > 
> > 
> > sparse warnings: (new ones prefixed by >>)
> > 
> > >> drivers/mtd/nand/nand_timings.c:45:29: sparse: constant 250000000000 is so big it is long
> > 
> > vim +45 drivers/mtd/nand/nand_timings.c
> > 
> >     29			.tCLS_min = 50000,
> >     30			.tCOH_min = 0,
> >     31			.tCS_min = 70000,
> >     32			.tDH_min = 20000,
> >     33			.tDS_min = 40000,
> >     34			.tFEAT_max = 1000000,
> >     35			.tIR_min = 10000,
> >     36			.tITC_max = 1000000,
> >     37			.tRC_min = 100000,
> >     38			.tREA_max = 40000,
> >     39			.tREH_min = 30000,
> >     40			.tRHOH_min = 0,
> >     41			.tRHW_min = 200000,
> >     42			.tRHZ_max = 200000,
> >     43			.tRLOH_min = 0,
> >     44			.tRP_min = 50000,
> >   > 45			.tRST_max = 250000000000,
> 
> Appending ULL to the numerical constant should fix this warning.
> 
> Do you want me to send a patch fixing this bug ?

No bug here, AFAICT. I noted already that, although this may appear to
overflow, it is in fact using standard behavior which says the constant
will be promoted to the proper type.

My comments:

  http://lists.infradead.org/pipermail/linux-arm-kernel/2014-April/252494.html

Looking at ISO draft N1124 (this is approximately C11, the first PDF I
had lying around), section 6.4.4.1 (statement 5):

  "The type of an integer constant is the first of the corresponding
  list in which its value can be represented."

So I think this is a sparse false positive. Notably, Fengguang's build
system also showed some complaints from an avr32 GCC toolchain. This is
also a bug.

Given that there are at least a few non-standards-compliant warnings out
there (and possibly real toolchain bugs, if they're giving these
warnings), perhaps it's prudent to patch this with a 'ULL' suffix, even
though I prefer not to work around bugs in tools in principle.

Brian