(FACT-654) Don't try to resolve Ldom facts on non-SPARC systems

[whopper]

In Solaris 11.2, Oracle introduced a rewrite of virtinfo which
supports both x86 and SPARC, whereas the old version only supported
SPARC. Since Ldoms are a SPARC-only feature, we now need to confine
the ldom facts to only resolve on SPARC based platforms.

Note that this information comes from community member Kristina Tripp,
who provided the necessary details on the new virtinfo implementation.

[whopper]

Note: though community members have mentioned that the behavior of virtinfo in SPARC systems should not have changed, we'd like to test this change out on a SPARC Solaris 11.2 system before merging.

[puppetcla]

CLA signed by all contributors.

[kylog]

Nit: maybe split onto multiple lines, e.g. so it shows up w/o right scrolling in github :)

[kylog]

Definitely test on an 11.2 sparc machine, but beyond that 👍

[enpointe]

Tested this out on a few Sparc setup running 11.2 (Sparc Virtual, T2000, and T4 LDOM). No problems seen

[enpointe]

bundler exec rspec test on Sparc bare metal passes with

1673 examples, 0 failures, 4 pending

I have seen one error message being generated when running facter now twice on sparc hardware. Not sure whether this is related to this change or not. I'm not seeing this message on facter 2.1.0

Timed out after 6 seconds while resolving fact='virtual', resolution=''

[whopper]

@enpointe thanks a ton for running this through on SPARC! As for that 'virtual' fact error, could we have you run facter --trace on that machine and paste what you see? My best bet is that the error is coming from the SunOS confined 'virtual' fact, which has a timeout of 6 seconds built in.

It looks like that change was actually added back in February, but was lost from our codebase until we merged our Facter 2 branch back into master this Summer.

[enpointe]

I can't reproduce that complaint with any regularity. I'll see if I can reproduce it though. Like I said I've only seen it twice but I thought I should mention it as I saw that error message was getting generated by core/resolvable.rb.
BTW, 'bundle exec rspec' tests aren't clean on a LDOM. There's obviously some additional work needed for facter on a sparc LDOM. The tests though are clean when run on a bare SPARC hardware and x86 hardware.

[whopper]

Interesting! Thanks for the heads up. We're hoping to get some testing going on in SPARC and in LDOMs soon to sort these sorts of things out. Thanks again for your help!

[whopper]

I'm closing this out to retarget the change at stable.