Add missing facts for Fedora 20, 21, and 26

[blackknight36]

Added fact files for Fedora 20, 21, and 26 to the facter 2.4 database.

[coveralls]

Coverage remained the same at 73.529% when pulling f568b6a on blackknight36:fedora_facts into ba64537 on camptocamp:master.

[DavidS]

Thank you very much for your contribution!

[blackknight36]

No problem. I have noticed a slight issue on modules that only have Fedora 20 listed as a supported OS. For example, rspec will show an error like this.

No facts were found in the FacterDB for: [{:facterversion=>"/\A2\.4\./", :operatingsystem=>"Fedora", :operatingsystemrelease=>"/^20/", :hardwaremodel=>"x86_64"}]

[DavidS]

@blackknight36 yeah, there seems to be something wrong with the facts you collected. https://github.com/camptocamp/facterdb/pull/50/files#diff-3f46c0bba56680b5787d4de77b8a8d98R16 shows "facterversion": "1.7.6" in the facts/2.4/fedora-20-x86_64.facts file. That's not right.

[blackknight36]

@DavidS It looks like the node itself has an older version of facter than the system I'm running tests on.

[root@node ~]# facter --version
1.7.6

Changing the facterversion fact to "2.4.1" does allow tests to run under Fedora 20 as shown below.

ipmi
on fedora-20-x86_64
should contain Package[OpenIPMI] with ensure => "installed"
should contain Package[ipmiutil] with ensure => "installed"
should contain Package[ipmitool] with ensure => "installed"
should contain File[ipmi_conf] that requires Package[OpenIPMI], Package[ipmiutil] and Package[ipmitool]
on centos-7-x86_64
should contain Package[OpenIPMI] with ensure => "installed"
should contain Package[ipmiutil] with ensure => "installed"
should contain Package[ipmitool] with ensure => "installed"
should contain File[ipmi_conf] that requires Package[OpenIPMI], Package[ipmiutil] and Package[ipmitool]

I can submit another PR to fix this if you want.

[DavidS]

The point of the facterdb is to collect the actual facts from a specific facter version on a specific OS. Just changing the facterversion in the json is not sufficient, since different facter versions have different sets of facts, and would make tests of other modules based on those facts invalid. Even for your own use-case, testing against 2.4 facts is pretty useless, if the code will run against 1.7.6 in production.

[DavidS]

I've opened #54 to rectify the bigger issue of wrong data.