Facter 4 compatibilty

[ekohl]

Since Facter 4 was released, it appears the selinux_python_command.rb fact has started to fail when rpm is not present. This happens when running tests on a non-Red Hat OS where rpm is not present:

Error while resolving custom fact fact='selinux_python_command', resolution='': Could not execute 'rpm -q python3-libsemanage'

It looks like Facter::Core::Execution.execute now raises an exception when the command fails where previously it didn't.

[ekohl]

Looks like this was also an issue in early Facter 2: https://tickets.puppetlabs.com/browse/FACT-409

[alexjfisher]

ping @IrimieBogdan

[blackknight36]

Are you certain that the confine statement is working? Seems like the exec statement shouldn't be evaluated at all on non-RedHat nodes although you would also run into issues on a RedHat based OS where rpm is not available.

[ekohl]

Note that this is in CI, which is always a bit different. I suspect the RH facts are loaded on an Ubuntu machine. That might mess up things.

[BogdanIrimie]

@ekohl thanks for the info, in the next release of Facter 4 (probably tomorrow) this will be fixed. Facter::Core::Execution.execute will no longer throw exception and will log a message if it failed to execute command.

[ekohl]

Great to hear. I'll revert theforeman/puppet-pulpcore@011f84b to verify it's been fixed.

[BogdanIrimie]

@ekohl We have identified the problem, when running Facter 4 in a gem context, the exit code should be 0, even if there were errors. There is an experimental fix, but it will not get in today's release. You can expect the fix to get in early next week.

[ekohl]

Thanks for the update! theforeman/puppet-pulpcore#83 should test that branch.

[ekohl]

It looks like that passes, but still outputs the error. Now that makes me wonder: were custom facts always executed in CI?

[BogdanIrimie]

@ekohl a new release of Facter 4 is out https://rubygems.org/gems/facter/versions/4.0.13 Please give it a try.

[ekohl]

Closing due to the new facter version.