Re: [PATCH V1 1/1] iio: Added Capella cm3232 ambient light sensor driver.

[Jonathan Cameron <jic23-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>]

On 05/01/15 17:56, Joe Perches wrote:
> On Mon, 2015-01-05 at 19:50 +0200, Daniel Baluta wrote:
>> On Mon, Jan 5, 2015 at 6:42 PM, Joe Perches <joe-6d6DIl74uiNBDgjK7y7TUQ@public.gmane.org> wrote:
>>> On Mon, 2015-01-05 at 16:20 +0200, Daniel Baluta wrote:
>>>> On Mon, Jan 5, 2015 at 3:09 PM, Joe Perches <joe-6d6DIl74uiNBDgjK7y7TUQ@public.gmane.org> wrote:
>>>>> On Mon, 2015-01-05 at 12:51 +0200, Daniel Baluta wrote:
>>>>>> On Thu, Jan 1, 2015 at 2:10 AM, Kevin Tsai <ktsai-GubuWUlQtMwciDkP5Hr2oA@public.gmane.org> wrote:
>>>>>>> CM3232 is an advanced ambient light sensor with I2C protocol interface.
>>>>>>> The I2C slave address is internally hardwired as 0x10 (7-bit).  Writing
>>>>>>> to configure register is byte mode, but reading ALS register requests to
>>>>>>> use word mode for 16-bit resolution.
>>> []
>>>>>> You could directly return i2c_smbus_write_byte_data(..).
>>>>>
>>>>> Sometimes it's better to return a specific value
>>>>> for the error instead of depending on correctness
>>>>> of all the indirect functions in the call chain.
>>>>>
>>>>> In this case, all the smbus_xfer functions must
>>>>> return 0 on success.  Do they?
>>>>
>>>> Yes.
>>>>
>>>> http://lxr.free-electrons.com/source/drivers/i2c/i2c-core.c#L2845
>>>
>>> This doesn't show that adapter->algo->smbus_xfer()
>>> returns 0, you have to look at the code for that
>>> indirectly called function.
>>
>> I based my answer on the comment at the top of the function:
>>
>> 2845  * This executes an SMBus protocol operation, and returns a negative
>> 2846  * errno code else zero on success.
> 
> Sure, but comments and code often differ and the
> implementation of any of those smbus_xfer functions
> could return a positive value like the byte value or
> the number of bytes written instead of 0.
> 
> For correctness, you'd have to inspect them all.
> 
> If some new future smbus_xfer function was written
> incorrectly, the return value from this function could
> now be positive.
> 
That would however, clearly be a bug and likely to cause all
sorts of issues elsewhere given the documentation requires that
it does return 0 on success and people will have been
relying on it.

Jonathan
> cheers, Joe
>