Re: [lm-sensors] [RFC PATCH 2/3] hwmon: sht15: add support for

From: Guenter Roeck <guenter.roeck@ericsson.com>
To: lm-sensors@vger.kernel.org
Subject: Re: [lm-sensors] [RFC PATCH 2/3] hwmon: sht15: add support for
Date: Sat, 26 Mar 2011 23:29:33 +0000	[thread overview]
Message-ID: <20110326232933.GA31800@ericsson.com> (raw)
In-Reply-To: <1300301050-15529-3-git-send-email-vivien.didelot@savoirfairelinux.com>

On Sat, Mar 26, 2011 at 01:16:00PM -0400, Jean Delvare wrote:
> On Tue, 22 Mar 2011 07:28:56 -0700, Guenter Roeck wrote:
> > On Tue, Mar 22, 2011 at 06:28:40AM -0400, Jonathan Cameron wrote:
> > > On 03/22/11 00:00, Vivien Didelot wrote:
> > > > Sure. After discussing about that, checksumming, otp_reload and
> > > > *_resolution attributes have been removed. battery_alarm will be splitted into
> > > > temp1_alarm and humidity1_alarm to respect the convention.
> > > Really, Guenter / Jean is the is the right option?  The alarm isn't really about
> > > either of the individual sensors, but rather about power to the chip?
> > 
> > I had suggested humidity1_alarm. Two alarms for one condition is overkill.
> > I find it better to have a standard attribute which a standard program such as
> > "sensors" may have a chance to display at some point. Jean may think differently - 
> > we did not discuss the matter.
> 
> Sorry for joining a little late in the game. If I understand properly

No problem - as always I appreciate your insight.

> the meaning of the "end of battery" status bit, humidity1_alarm is not
> a suitable name. The *_alarm attributes are for off-limit measurements,
> but the SHT1x has no limits in the first place. The "end of battery"
> status bit means "measurements may be invalid", so I would map it to
> humidity1_fault and temp1_fault. The later is already standardized, and
> it would be easy to introduce the former. The usual meaning of
> temp*_fault is a little different (broken / open / short sensor) but
> its standard definition is fairly generic.
> 
> When the fault flag is set, "sensors" will display "FAULT" instead of
> the measurement value, which is the right thing to do here if the
> measurements are unreliable.
> 
> And there is nothing wrong with mapping a single register bit to two
> attributes, if it makes sense as is the case here. I think we already
> did it in the past for other chips.
> 
Ok.

Thanks,
Guenter

_______________________________________________
lm-sensors mailing list
lm-sensors@lm-sensors.org
http://lists.lm-sensors.org/mailman/listinfo/lm-sensors