Wrong HighVoltageTransfer value on CyberPower PR1000ELCDRT2U

[LucidityCrash]

Just got a CyberPower PR1000ELCDRT2U and trying to make NUT work, I noticed that the High Voltage Transfer Value is incorrect
Running : /lib/nut/usbhid-ups -a cyberpower1 -DDDD the relevant section shows :
0.618726 string_to_path: depth = 3
0.618744 Report[buf]: (3 bytes) => 11 1a 01
0.618760 Path: UPS.Input.HighVoltageTransfer, Type: Feature, ReportID: 0x11, Offset: 0, Size: 16, Value: 90

The Hex is correct 011a = 282 but the value reported shows as 90

The Low Voltage transfer is correct :

0.618674 string_to_path: depth = 3
0.618689 Report[buf]: (2 bytes) => 10 aa
0.618705 Path: UPS.Input.LowVoltageTransfer, Type: Feature, ReportID: 0x10, Offset: 0, Size: 8, Value: 170

I've been looking through the code but I'm struggling to see where the value is set from the raw hex, could I get some pointers ?

[clepple]

This is probably the same underlying issue as in #482, just with a different set of HID parameters left open on the HID parsing stack. Think of it as the HID equivalent of mismatched parentheses. If you post the value of the HID Report Descriptor, I can explain a little more of what is going on. I forget what debug level shows that descriptor, but it's the wall of hex at either 4 or 5 -Ds.

HIDGetDataValue() calls logical_to_physical() on the raw hex value from the report. These are both in libhid.c.

I don't think we currently have code to set high/low transfer on CyberPower units (#438) so this is basically a display issue at this point. So you can paper over the problem with an override.* directive. To solve the underlying HID descriptor problem, we need a way to replace HID Report Descriptors: #169

[jimklimov]

Cheers, per clues here and there, this issue may be related to the fixes that landed into NUT v2.8.0 release (and addressed more in "master" branch afterwards). Do you have a chance to test any of those builds - whether NUT compatibility with the device has improved?

[jimklimov]

PR #1245 is presumed to have fixed the broken CPS high voltage transfer reports discussed in this issue. Both high and low values were reported reasonable in #1512 and #2718 testing, though possibly on models too different to this issue?...