gcc 4.9 build warnings (was: Re: arm-soc build: 2917 warnings 0 failures (arm-soc/v3.18-rc1-20-g06c0773))

[Thierry Reding]

On Fri, Oct 24, 2014 at 03:41:48PM +0200, Ard Biesheuvel wrote:
> On 24 October 2014 15:27, Arnd Bergmann <arnd at arndb.de> wrote:
> > On Friday 24 October 2014 13:50:05 Ard Biesheuvel wrote:
> >> On 24 October 2014 13:37, Arnd Bergmann <arnd at arndb.de> wrote:
> >> > On Friday 24 October 2014 12:59:40 Ard Biesheuvel wrote:
> >> >> On 24 October 2014 12:52, Russell King - ARM Linux
> >> >> <linux at arm.linux.org.uk> wrote:
> >> >> > On Fri, Oct 24, 2014 at 12:49:50PM +0200, Arnd Bergmann wrote:
> >> >> >> >       1 drivers/video/fbdev/sm501fb.c:245:2: warning: format '%zd' expects argument of type 'signed size_t', but argument 8 has type 'size_t' [-Wformat=]
> >> >> >> >       1 mm/percpu.c:895:3: warning: format '%zu' expects argument of type 'size_t', but argument 2 has type 'unsigned int' [-Wformat=]
> >> >> >> >       1 mm/percpu.c:895:3: warning: format '%zu' expects argument of type 'size_t', but argument 3 has type 'unsigned int' [-Wformat=]
> >> >> >>
> >> >> >> The warnings are completely unchanged, still 249 unique warnings involving size_t,
> >> >> >> using this patch:
> >> >> >
> >> >> > That's probably because the compiler is expecting size_t to be typedef'd
> >> >> > to __SIZE_TYPE__ and isn't expecting anyone to change it.
> >> >> >
> >> >>
> >> >> Indeed, I wouldn't expect the printf format validation code inside GCC
> >> >> to care about the actual values of macros and typedefs.
> >> >> Could someone dump the builtin #define's of that compiler? For
> >> >> instance, my bare metal 4.9 BE GCC gives me
> >> >>
> >> >> $ /usr/local/gcc-linaro-armeb-none-eabi-4.9-2014.06_linux/bin/armeb-none-eabi-gcc
> >> >> -E -dM - <<<"" |grep _TYPE__
> >> >> #define __UINT_LEAST8_TYPE__ unsigned char
> >> >> #define __SIG_ATOMIC_TYPE__ int
> >> >> #define __UINTMAX_TYPE__ long long unsigned int
> >> >> #define __INT_FAST16_TYPE__ int
> >> >> #define __INT_FAST64_TYPE__ long long int
> >> >> #define __UINT8_TYPE__ unsigned char
> >> >> #define __INT_FAST32_TYPE__ int
> >> >> #define __UINT_LEAST16_TYPE__ short unsigned int
> >> >> #define __SIZE_TYPE__ unsigned int
> >> >> #define __INT8_TYPE__ signed char
> >> >> #define __INT_LEAST16_TYPE__ short int
> >> >> #define __UINT_LEAST64_TYPE__ long long unsigned int
> >> >> #define __UINT_FAST16_TYPE__ unsigned int
> >> >> #define __CHAR16_TYPE__ short unsigned int
> >> >> #define __INT_LEAST64_TYPE__ long long int
> >> >> #define __INT16_TYPE__ short int
> >> >> #define __INT_LEAST8_TYPE__ signed char
> >> >> #define __INTPTR_TYPE__ int
> >> >> #define __UINT16_TYPE__ short unsigned int
> >> >> #define __WCHAR_TYPE__ unsigned int
> >> >> #define __UINT_FAST64_TYPE__ long long unsigned int
> >> >> #define __INT64_TYPE__ long long int
> >> >> #define __WINT_TYPE__ unsigned int
> >> >> #define __UINT_LEAST32_TYPE__ long unsigned int
> >> >> #define __INT_LEAST32_TYPE__ long int
> >> >> #define __UINT64_TYPE__ long long unsigned int
> >> >> #define __INT_FAST8_TYPE__ int
> >> >> #define __UINT_FAST32_TYPE__ unsigned int
> >> >> #define __CHAR32_TYPE__ long unsigned int
> >> >> #define __INT32_TYPE__ long int
> >> >> #define __INTMAX_TYPE__ long long int
> >> >> #define __PTRDIFF_TYPE__ int
> >> >> #define __UINT32_TYPE__ long unsigned int
> >> >> #define __UINTPTR_TYPE__ unsigned int
> >> >> #define __UINT_FAST8_TYPE__ unsigned int
> >> >>
> >> >> What is surprising here is that __SIZE_TYPE__ is int not long. Could
> >> >> we in fact be dealing with a 4.9 bare metal GCC bug here?
> >> >
> >> > This is what I get on every arm gcc version on my system,
> >> > all glibc or uClibc targetted, compared to yours:
> >> >
> >> >  #define __UINT_FAST64_TYPE__ long long unsigned int
> >> >  #define __INT64_TYPE__ long long int
> >> >  #define __WINT_TYPE__ unsigned int
> >> > -#define __UINT_LEAST32_TYPE__ long unsigned int
> >> > -#define __INT_LEAST32_TYPE__ long int
> >> > +#define __UINT_LEAST32_TYPE__ unsigned int
> >> > +#define __INT_LEAST32_TYPE__ int
> >> >  #define __UINT64_TYPE__ long long unsigned int
> >> > -#define __INT_FAST8_TYPE__ int
> >> > +#define __INT_FAST8_TYPE__ signed char
> >> >  #define __UINT_FAST32_TYPE__ unsigned int
> >> > -#define __CHAR32_TYPE__ long unsigned int
> >> > -#define __INT32_TYPE__ long int
> >> > +#define __CHAR32_TYPE__ unsigned int
> >> > +#define __INT32_TYPE__ int
> >> >  #define __INTMAX_TYPE__ long long int
> >> >  #define __PTRDIFF_TYPE__ int
> >> > -#define __UINT32_TYPE__ long unsigned int
> >> > +#define __UINT32_TYPE__ unsigned int
> >> >  #define __UINTPTR_TYPE__ unsigned int
> >> > -#define __UINT_FAST8_TYPE__ unsigned int
> >> > +#define __UINT_FAST8_TYPE__ unsigned char
> >> >
> >>
> >> So no __SiZE_TYPE__ then? That's surprising ...
> >>
> >
> > Sorry for being unclear here: this is only the diff between my version and
> > yours, so everything that is not listed above __UINT_FAST64_TYPE__ is the
> > same on both compilers.
> >
> 
> Ah, ok. So apparently, size_t is not ambiguous between bare metal and
> glibc GCC, so we are looking at something else here.

GCC complains about the format specifier being wrong. %zu/%zd are the
correct specifiers for variables of type size_t/ssize_t, so wherever a
size_t or ssize_t is used as parameter it should have a corresponding
%zu or %zd specifier.

Why not just fix it properly instead of mucking about with the size_t
typedef?

Thierry
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20141106/4058cb8f/attachment.sig>