2236 nut scanner reports non existent driver for nut discovery

[aquette]

When discovering NUT, either with Avahi or old method, the reported driver is "nutclient", which doesn't (yet) exists.
This should be fixed to dummy-ups, which will use the repeater mode, while waiting for #2232

the short term fix is to symlink nutclient to dummy-ups (example on Debian):

cd /lib/nut && ln -s dummy-ups nutclient

[aquette]

seriously @jimklimov?
https://ci.networkupstools.org/job/nut/job/nut/job/PR-2245/1/pipeline-console/?selected-node=252

please put such targets as optional, and revise the CI to be dev friendly. Breaking the momentum for such errors is no good ;)

[jimklimov]

Can't see details too well on mobile on the road, but errors look like broken automake syntax? Did you change the makefiles to match?

[aquette]

boils down to 1 single poor OpenIndiana failure among bagazillions working, and that destroys the whole castle :/

configure.ac:2739: error: possibly undefined macro: AC_DISABLE_STATIC

Please fix the CI @jimklimov, spasibo :)

[jimklimov]

Well, we sort of (officially) pride ourselves on supporting builds of NUT everywhere it could be built before, hence the very diverse CI-testing matrix with C99+ and 2000'ish or newer (du'jour and beyonder) toolkit implenentations. While not part of regular builds, there were in recent past questions about Solaris 8 and CentOS 6, and a few fixes to make those happen. Developer momentum can quite quickly happen on their few systems, but non-regression for everyone has to happen on all the 180-ish os/toolkit combos.

I'll try to check that code and error-report if there will be signal later in the night, or after holidays at worst. Maybe just some disk space ran out and diags were misread.

[aquette]

for sure a pride, I've been the one porting it for more than 15 years.
Still, that should block a final release, not a dev flow!
This CI thing of yours is so big, it's collapsing under its own weight!

I'm gonna force gates to move fast, we'll fix after ;)

[jimklimov]

As noted elsewhere, the issue was in the end due to NUT CI farm VM migration to a new provider, and it ended up pointing to a wrong container (not set up for builds) as the OI agent.