[Intel-gfx] [igt-dev] [PATCH i-g-t 2/7] Cast void * pointer used in arithmetic to uint32_t*

[Rodrigo Siqueira]

Hi,

Thanks for the feedback,

I'm just a little bit confused about the next step here. Is it worth to
fix this patch and the others? I got confused since you said that IGT
will remove autotools and some of the warnings here does not appear on
meson.

Best Regards
Rodrigo Siqueira

On 07/10, Arkadiusz Hiler wrote:
> On Tue, Jul 10, 2018 at 03:09:25PM +0100, Chris Wilson wrote:
> > Quoting Arkadiusz Hiler (2018-07-10 14:58:49)
> > > On Sat, Jul 07, 2018 at 08:22:38PM -0300, Rodrigo Siqueira wrote:
> > > > This commit fixes the GCC warning:
> > > > 
> > > > warning: pointer of type ‘void *’ used in arithmetic [-Wpointer-arith]
> > > >      memset(ptr + offsets[0], full_range ? 0x00 : 0x10,
> > > >                 ^
> > > > warning: pointer of type ‘void *’ used in arithmetic [-Wpointer-arith]
> > > >      memset(ptr + offsets[1], 0x80,
> > > > 
> > > > This commit cast the ptr pointer, which is a void *, to uint32_t * in
> > > > the pointer arithmetic operation.
> > > 
> > > This will change semantics, as according to GNU C standard[1], void
> > > pointers have a size of 1 for all arithmetical purposes.
> > > 
> > > So you should be using uint8_t (or char) pointer instead.
> > 
> > Please just fix the compiler flags, we want close compatibility with the
> > kernel coding standards which explicitly allow void arithmetic for the
> > simplicity it lends to writing code.
> > -Chris
> 
> Fair point.
> 
> We don't rise the error with meson, so it is not a change in the gcc
> defaults. Somehow autotools manage to end up adding -Wpointer-arith to
> BASE_CFLAGS.
> 
> I don't think we should invest time into making autotools behave, since
> it's going to be dropped completely. Hopefully this will happen sooner
> than later.
> 
> -- 
> Cheers,
> Arek