[PATCH:xf86-video-mga 2/2] Use correct type in function call.

[Mark Kettenis]

>  On Mon, Feb 23, 2015 at 4:28 AM, Thomas Klausner <wiz at netbsd.org> wrote:
> > Fixes
> > passing argument 2 of 'pci_device_cfg_read_u32' from incompatible
> > pointer type
> > pciaccess.h:153:5: note: expected '__uint32_t *' but argument is of type
> > 'CARD32 *'
> >
> > Signed-off-by: Thomas Klausner <wiz at NetBSD.org>
> > ---
> >  src/mga_dac3026.c | 6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
> >
> > diff --git a/src/mga_dac3026.c b/src/mga_dac3026.c
> > index 97c2c7a..e825702 100644
> > --- a/src/mga_dac3026.c
> > +++ b/src/mga_dac3026.c
> > @@ -852,8 +852,12 @@ MGA3026Save(ScrnInfoPtr pScrn, vgaRegPtr vgaReg,
> > MGARegPtr mgaReg,
> >                 mgaReg->DacRegs[i]       = inTi3026(MGADACregs[i]);
> >
> >  #ifdef XSERVER_LIBPCIACCESS
> > -       pci_device_cfg_read_u32(pMga->PciInfo, & mgaReg->Option,
> > +    {
> > +       uint32_t Option;
> > +       pci_device_cfg_read_u32(pMga->PciInfo, & Option,
> >                                 PCI_OPTION_REG);
> > +        mgaReg->Option = Option;
>
>  I don't understand why the previous code generated that warning (isn't
>  __uint32_t* implicitly castable to CARD32*?)

On 32-bit systems, CARD32 is "unsigned long", whereas uint32_t is usually
"unsigned int".  And since "int" and "long" might differ in size on other
systems, the compiler (rightfully) complains.

>  but even so, why isn't a cast the appropriate solution?

Because casts almost never are the appropriate solution if they can be
avoided?