libshut.c riello_ser.c nut-scanner bits: banish non-standard u_char and u_short

[jimklimov]

Probably standard C99 "uint8_t" and 16(?) bit short would be more proper,
especially for serial-line protocols. But the original code used types
derived from hardware implementation sized "char" and "short" - so they
stayed this way here.

Stems from efforts on #823 and #844 to get more clean and green system setups in CI. Due to the comment above, review would be welcome.

[aquette]

Seems indeed good, with a 5 seconds review. Testable for non reg on any Eaton unit with a serial port and having USB HID (ie not older power ware with XCP)

[clepple]

LGTM

[jimklimov]

Thanks @aquette for the suggestion, now I intend to test this on an OmniOS box that has Eaton 9PX 5000i UPS attached by USB cable as well as seen over SNMP :) Not sure how much the USB drivers would help with bug-hunt in serial port code, but at least similar variable type change in SNMP code is more auditable.

Not sure what to look out for, probably diff's of outputs from upsc with one or other same-named driver build.

[aquette]

testing through USB won't give anything, since libshut is the equivalent marshaling layer of libusb.
Then it's the common HID core above.
AFAIR, 9PX also have a serial port (RS 232 / DB9). Keep in mind that P2P ports are mutually exclusive.
So when testing RS232, remove the USB cord from the UPS ;)
hope it helps, I've no more ups@home, so can't help further

[jimklimov]

In this case, my physical setup that is stranded in the now-inaccessible office won't be of much help.

I looked at systems headers I could get at, everywhere that has an u_char defined it is an unsigned char (same for shorts), so this PR does not change the outcome for those.

I suppose a further pedantic change to turn such arch-dependent cases into uint8_t and uint16_trespectively, would need more testing with real hardware.