Follow-up Comment #17, sr #111312 (group freeipmi): So that's pretty interesting. It suggests the error is in ipmi_sensor_decode_value(). i.e. the actual mathematical conversion from the raw byte to number.
When I compared my code to ipmitool's, there is a minor computational
difference, where I cast an integer to a double BEFORE a mathematical
calculation. They cast to the double AFTER the calculation. i.e. i do
result = (double)a * (double)b
and they do
result = (double) (a * b)
I'm wondering if this could be some very rare precision loss. Something that
shouldn't be a problem on 99.9% of cases, but somehow you hit it.
I will send you another debug patch tomorrow and we'll see how things are
computed.
If there ends up being a computational difference in there, I will have to dig
into the IPMI spec and see what they really define in there.
_______________________________________________________
Reply to this item at:
<https://savannah.gnu.org/support/?111312>
_______________________________________________
Message sent via Savannah
https://savannah.gnu.org/
signature.asc
Description: PGP signature
