Re: [PATCH v2 3/4] iio: adc: ti-ads7950: switch to using guard() notation

[David Lechner]

On 2/22/26 3:31 PM, Dmitry Torokhov wrote:
> On Sat, Feb 21, 2026 at 11:20:42AM -0600, David Lechner wrote:
>> On 2/19/26 1:51 AM, Andy Shevchenko wrote:
>>> On Wed, Feb 18, 2026 at 06:29:27PM -0800, Dmitry Torokhov wrote:
>>>> guard() notation allows early returns when encountering errors, making
>>>
>>> guard()()
>>>
>>> // strictly speaking
>>>
>>>> control flow more obvious. Use it.
>>>
>>> I like the change, but...
>>>
>>>> Also variables that now only hold error codes (or 0) are renamed to
>>>> "error" to make their purpose clearer.
>>
>> Normally I would not give my opinion on this, but since I wrote the driver
>> originally, I will say please don't rename. I prefer to always use "ret".
> 
> I hope I can convince you otherwise.

I'm afraid not. "ret" is used > 35k times in IIO and err is used < 3k times
and error < 1k times. So I am really used to seeing only "ret" for errors and
"error" looks very unusual to my eyes because of this.

You mentioned valuing the common pattern in your other reply, so I hope
you can understand that this is what I value more here.

> 
> IMO "ret" or "retval" should be used when the returned value is intended
> to be used during normal operation. For cases where we only expect to
> have an error or 0 for success "error" or "err" is more appropriate.
> This allows you to write
> 
> 	error = do_action(...);
> 	if (error) {
> 		// handle error somehow, typically simply report.
> 	}
> 
> This also helps when reading the code as you know that there is usually
> no reason to care about the specific value in this variable (maybe
> except -EPROBE_DEFER).
> 
> I will push the conversion ret -> error to the very last patch so it can
> easily be dropped if I am unsuccessful in swaying your opinion.
> 
> Thanks.
> 

To give you fair warning, it will still be NAK from me.