> Well, looking at the conversion formula again, this idea is not 100 % > perfect, because it doesn't care to scale the offset value. Currently, > the offset is 0 but there is no guarantee, that it will always be 0. This isn't a problem. The formulae are applied *after* libsensors (or any other program such as gkrellm) reads the value from sysfs or procfs. For example, if you have a raw reading of 1.8V, in 2.4 it'll be stored internally as 180, and read through procfs as 1.80. In 2.6 it'll be stored internally as 1800 and exported through sysfs as 1800 too. But libsensors knows about magnitudes in both cases and will read the value as 1.80. And any formula in sensors.conf will be applied to *this* value. So, as you can see, it is safe to have a single formula for both versions. > What do you think about some further RW files > > in_refvoltage = 0x21 > in_offset0 = 0x00 > in_offset1 = 0x00 > in_offset2 = 0x00 > in_multiplier0 = 0x31 > in_multiplier1 = 0x14 > in_multiplier2 = 0x0a > > that initally have the values of the first driver release, which were > taken from dmidecode. The driver would then use those values and > supply scaled values for in0, in1 and in2. I find it more confusing than helpful. All drivers have their formula in sensors.conf and I believe we should stick to that model. Making something different for one driver will make third-party apps unable to handle it correctly unless they know the trick and add chip-specific code. Remember that our aim with 2.6 and sysfs is to be able to live without any specific code. Readings from sysfs and sensors.conf should be sufficent to read temperatures, voltages and fans values and configure limits and such, with no additional details given. It's still not perfect since a few things are not well standardized (think to alarms for example) but that's the idea. So I suggest you go with your formulas in sensors.conf.eg, they look just fine to me. > BTW: was the multiplier for temperatures changed from 100 to 1000 > around 2.6.0-test5, because gkrellm-2.1.25 uses 100 instead of 1000? I can't remember exactly when the change was made. I remember it was discussed and we opted for something standard enough so that it would be OK for everyone. More likely, gkrellm changed to accomodate our decision. -- Jean Delvare http://www.ensicaen.ismra.fr/~delvare/
