On ma, 2014-06-16 at 12:52 -0700, Troy Kisky wrote: > On 6/16/2014 12:41 PM, Jeroen Hofstee wrote: > > Since &rgb2ycbcr_coeff and friends are declared const, but > > assigned to a void pointer, clang will warn that the const > > is implicity casted away. Add const to indicate the pointer > > points to const data to get rid of these warnings. > > > > Cc: Stefano Babic <sba...@denx.de> > > Signed-off-by: Jeroen Hofstee <jer...@myspectrum.nl> > > --- > > drivers/video/ipu_disp.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/drivers/video/ipu_disp.c b/drivers/video/ipu_disp.c > > index cefd2dc..31b5a10 100644 > > --- a/drivers/video/ipu_disp.c > > +++ b/drivers/video/ipu_disp.c > > @@ -33,7 +33,7 @@ enum csc_type_t { > > > > struct dp_csc_param_t { > > int mode; > > - void *coeff; > > + void * const coeff; > > }; > > > > #define SYNC_WAVE 0 > > > > Doesn't this says the pointer is constant, not that it points to constant > data? > So, commit log needs updated or > const void *coeff;
The reason I wrongly changed it from void const * to void * const was to circumvent gcc complaining with: drivers/video/ipu_disp.c: In function ‘ipu_dp_csc_setup’: drivers/video/ipu_disp.c:393:8: warning: assignment discards ‘const’ qualifier from pointer target type [enabled by default] This can be solved with: @@ -390,7 +390,7 @@ void ipu_dp_csc_setup(int dp, struct dp_csc_param_t dp_csc_param, __raw_writel(reg, DP_COM_CONF()); } - coeff = dp_csc_param.coeff; + coeff = (const int (*)[5][3]) dp_csc_param.coeff; if (coeff) { __raw_writel(mask_a((*coeff)[0][0]) | But I am not too happy with that. Happen to know a better way? Regards, Jeroen _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot