On Dienstag, 14. Februar 2012, Jonathan Cameron wrote:
> On 2/14/2012 3:04 PM, Dan Carpenter wrote:
> > On Tue, Feb 14, 2012 at 03:35:36PM +0100, Danny Kukawka wrote:
> >> Fix for:
> >> drivers/staging/iio/gyro/adis16080_core.c: In function
> >> ‘adis16080_read_raw’:
> >> drivers/staging/iio/gyro/adis16080_core.c:99:8: warning: ‘ut’
> >> may be used uninitialized in this function [-Wuninitialized]
> >>
> >> Initialize ut and change error handling from adis16080_read_raw().
> >>
> >> Signed-off-by: Danny Kukawka<danny.kukawka@xxxxxxxxx>
> >> ---
> >> drivers/staging/iio/gyro/adis16080_core.c | 4 ++--
> >> 1 files changed, 2 insertions(+), 2 deletions(-)
> >>
> >> diff --git a/drivers/staging/iio/gyro/adis16080_core.c
> >> b/drivers/staging/iio/gyro/adis16080_core.c index 1815490..e0b2a29
> >> 100644
> >> --- a/drivers/staging/iio/gyro/adis16080_core.c
> >> +++ b/drivers/staging/iio/gyro/adis16080_core.c
> >> @@ -82,7 +82,7 @@ static int adis16080_read_raw(struct iio_dev
> >> *indio_dev, long mask)
> >> {
> >> int ret = -EINVAL;
> >> - u16 ut;
> >> + u16 ut = 0;
> >> /* Take the iio_dev status lock */
> >>
> >> mutex_lock(&indio_dev->mlock);
> >> @@ -94,7 +94,7 @@ static int adis16080_read_raw(struct iio_dev
> >> *indio_dev, if (ret< 0)
> >> break;
> >> ret = adis16080_spi_read(indio_dev,&ut);
> >> - if (ret< 0)
> >> + if (ret)
> >> break;
> >
> > Either one of these changes would silence the warning from gcc
> > (which is a false positive). I would keep the "ut = 0;" change and
> > leave the error handling the same. That way we check for less than
> > zero consistently instead of checking some for non-zero and some for
> > less than zero.
>
> Agreed. I've always been lazy on this one as it was a false positive,
> but might as well
> get rid of it...
I'm okay with 'ut = 0;'.
I guess it's hard for gcc to find out that spi_read() will only return 0 or a
negative value in error case. That's why gcc may assume 'ut' could be used
initialized since adis16080_spi_read() may return with a value > 0.
And in case spi_read() or one of the functions behind would change it could
cause an initialized 'ut'. The code may should always check for "ret != 0" as
already done in some other places (in similar cases) to be save.
Danny
--
To unsubscribe from this list: send the line "unsubscribe linux-iio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
