On 01/31/2017 02:00 PM, Mark Brown wrote:
On Thu, Jan 26, 2017 at 02:09:42PM -0600, Pierre-Louis Bossart wrote:
Optional gpio handling should not cause an error status and prevent probing. Remove error return and move error message to dev_info @@ -3660,8 +3660,7 @@ static int rt5645_i2c_probe(struct i2c_client *i2c, GPIOD_IN);
if (IS_ERR(rt5645->gpiod_hp_det)) {
dev_err(&i2c->dev, "failed to initialize gpiod\n");return PTR_ERR(rt5645->gpiod_hp_det);
dev_info(&i2c->dev, "failed to initialize gpiod\n");I'd expect this to depend on what the error is - the normal pattern is to check for whatever the error code for "this doesn't exist" is and ignore that while treating other things (especially -EPROBE_DEFER but also other things) as errors.
yes, I'll apply the same filter on -ENOENT as for the test for the mclock last month. I should have thought about it, thanks for pointing this out.