Previously this was incorrectly handling facts that were of the form
foo=1+1=2 due to the ='s in the actual fact contents. Fix this and
add tests to try and prevent regressions.
Without this patch Puppet Enterprise users who install the most recent
version of stdlib lose the ability to resolve certain facts critical to
the operation of Puppet Enterprise. These facts are defined externally
in the file
`/etc/puppetlabs/facter/facts.d/puppet_enterprise_installer.txt`.
As an example, Puppet Enterprise catalogs fail to compile if the
`fact_stomp_server`, and `fact_stomp_port` facts are not defined.
`facter_dot_d` was removed from stdlib version 4 because Facter version
1.7 now supports external facts defined in
`/etc/puppetlabs/facter/facts.d/puppet_enterprise_installer.txt`.
Puppet Enterprise does not yet include Facter 1.7, however. The most
recent PE release, 2.8.1, includes Facter 1.6.17. With this version of
Facter, users who replace the version of stdlib that ships with PE with
the most recent version from the Forge will lose the ability to resolve
facts from
`/etc/puppetlabs/facter/facts.d/puppet_enterprise_installer.txt`.
This patch addresses the problem by detecting if Facter version < 1.7 is
loaded. If so, then the facter_dot_d.rb facts will be defined using the
stdlib custom fact. If Facter >= 1.7 is being used then stdlib will not
define external facts.
This reverts commit 8fc00ea5b6.
We're restoring facts_dot_d support to stdlib because users are pulling
in the latest version of stdlib while on Puppet Enterprise which breaks
the operation of PE itself when the fact_stomp_server and
fact_stomp_port facts are not defined. They are not defined in PE
because PE runs with Facter 1.6.17 and Puppet 2.7.21
This reverts commit d6d23b495c.
Why? Because this change set should actually be in master and our
merge-up process reverted the change set in master when I reverted from
2.4.x.
This patch reverts the revert, restoring the original change set.
This reverts commit 74e6411157, reversing
changes made to 417d219aa6.
Here's why:
Actually... I just screwed this up.
I merged this new fact into 2.4.x but it's not fixing any bug. It's adding a
new fact, so this should go into master and we should release 2.5 since this is
new, backwards-compatible functionality.
* 2.4.x:
Prevent undefined method `split' for nil:NilClass with pe_foo_version facts
(maint) Clear all facts before each example
Add spec tests for pe_version facts
Add PE facts to stdlib
Conflicts:
spec/spec_helper.rb
Without this patch the pe_major_version, pe_minor_version, and
pe_patch_version facts directly depend on the pe_version fact in a
manner that calls split directly on the return value.
This is a problem because Fact values are not always guaranteed to
return strings, or objects that respond to split. This patch is a
defensive measure to ensure we're always calling the split method on a
string object.
If the Fact returns nil, this will be converted to an empty string
responding to split.
As many PE modules have PE specific functionality, but are deployed to all
nodes, including FOSS nodes, it is valuable to be able to selectively enable
those PE specific functions. These facts allow modules to use the is_pe fact to
determine whether the module should be used or not. The facts include is_pe,
pe_version, pe_major_version, pe_minor_version, and pe_patch_version. For PE
2.6.0 those facts would have values true, 2.6.0, 2, 6, and 0, respectively.
This reverts commit cc414a422d, reversing
changes made to 29f8f89c19.
Conflicts:
README.markdown
Without this patch, there is no facts_dot_d functionality and we don't
have it implemented in Facter 2.0. This is a problem because Puppet
Enterprise and many users rely on facts.d support. We're also backwards
compatible with Facter 1.6 in stdlib 3.0 so this is a bug fix.
Since facts_dot_d will eventually be removed and replaced by
external facts, warn users who are using a ttl on their external
facts that this feature will not be in Facter external facts.
Provide a link to a page explaining how to cache fact values
without the ttl functionality.
Since external facts does the same thing as facts_dot_d (except
allow ttl for external facts and allow some Windows executable
external facts), and external facts is the prefered solution,
remove facts_dot_d in the next major relase of stdlib.
Puppet apply does not add the stdlib lib directory to the $LOAD_PATH.
This is a problem because the puppet_vardir fact requires the
puppet_settings library to be available for the `with_puppet` utility
method.
Without this patch, puppet apply will result in the following error:
$ puppet apply --modulepath=/vagrant/modules -e 'notice $puppet_vardir'
warning: Could not load fact file stdlib/lib/facter/puppet_vardir.rb: no such file to load -- facter/util/puppet_settings
notice: Scope(Class[main]):
notice: Finished catalog run in 0.01 seconds
With this patch applied, puppet apply works as expected:
$ puppet apply --modulepath=/vagrant/modules.pe -e 'notice $puppet_vardir'
notice: Scope(Class[main]): /Users/jeff/.puppet/var
notice: Finished catalog run in 0.01 seconds
This patch defensively tries to load facter/util/puppet_settings. If it cannot
load it, it falls back to trying to explicitly locate and load the library.
Once puppet is fixed such that a modules lib directory is truly in the
$LOAD_PATH, the fall back implementation will no longer be exercised since the
LoadError should not be raised.
On Windows, we have no folders that match up to the default set of
directories the facter_dot_d fact looks in by default. This is a
problem because the Puppet Enterprise installer writes out the following
facts by default, and our modules require them to be present:
% cat /etc/puppetlabs/facter/facts.d/puppet_enterprise_installer.txt
fact_stomp_port=61613
fact_stomp_server=puppetmaster
fact_is_puppetagent=true
fact_is_puppetmaster=true
fact_is_puppetconsole=true
On windows, the Puppet confdir is quite variable. On 2003 systems we
default to the All Users application data directory. On 2008 systems we
default to the ProgramData directory. The actual configuration
directory varies depending on the Puppet or Puppet Enterprise branding.
In order to simplify all of this variable behavior, this patch fixes the
problem by automatically looking for facts in
`%COMMON_APPDATA%/PuppetLabs/facter/facts.d`
This patch paves the way for the MSI installer to use an IniFile element
to write custom facts during installation.
Without this patch the PE modules don't have a way to identify a
filesystem path where it's OK to place variable data related to managing
the target node. This is a problem when a module like pe_compliance
needs to write a wrapper script to the node's filesystem.
This patch addresses the problem by exposing the node's Puppet[:vardir]
setting as a Facter fact.
This fact value will be set to `nil` if Puppet is not loaded into
memory. If Puppet is loaded, e.g. using `facter --puppet` or using
`puppet agent` or `puppet apply` then the fact will automatically set
the value to Puppet[:vardir]
The value of this setting is subject to Puppet's run_mode.
This patch implements a new utility method in the standard library
module named `Facter::Util::PuppetSettings.with_puppet`. The method
accepts a block and will only invoke the block if the Puppet library is
loaded into the Ruby process. If Puppet is not loaded, the method
always returns nil. This makes it easy to define Facter facts that only
give values if Puppet is loaded in memory.
Without this patch the root_home fact fails on windows. This patch
fixes the problem by only calling methods on the object returned by the
`getent passwd root` command if the object evaluates to true.
Because there is no root account on Windows the code block simply
returns `nil` which makes the Facter fact undefined on Windows
platforms.
The root cause of the failure is that we always expected the command to
succeed and return something useful, and it may not on all supported
platforms.
Without this patch an infinite loop will be entered if the json and
rubygems libraries are not available.
This patch fixes the problem by retrying the `require 'json'` only if
rubygems was successfully loaded for the first time. Subsequent
attempts to load rubygems will cause the LoadError exception from a
missing json library to be re-raised.
Thanks to Krzysztof Wilczynski for pointing out this issue.
Without this patch applied, the stdlib module does not provide a
root_home fact. This fact is necessary to easily determine the root
account home directory on platforms Puppet is supported on.
The major variations this fact address are:
---
solaris: /
linux: /root
macosx: /var/root
Spec tests using rspec have been provided as well to cover these three
general cases. Windows tests are marked as pending.
Based on feedback from Luke, the facts.d directory should at least match
the directory that will be used by Facter 2.0.
Reading #2157 I believe the Facter 2.0 facts.d feature is reasonably API
compatible with this custom fact from R.I. so I'm comfortable using the
same filesystem path.
Change in behavior: Now look for facts in:
* /etc/facter/facts.d
* /etc/puppetlabs/facter/facts.d
This fact is a direct copy of R.I.'s work at
https://github.com/ripienaar/facter-facts
This is necessary plumbing to allow the installer to write a simple text
file based on the role the node is receiving. For example:
$ cat /etc/puppetlabs/facts.d/puppet_enterprise_mcollective.txt
fact_stomp_port=61613
fact_stomp_server=puppetmaster
fact_is_puppetagent=true
fact_is_puppetmaster=true
The mcollectivepe module relies on these facts being set and we need a
persistent place to write them during the interview process and later
read them when puppet agent runs to configure MCollective on the agent
systems.
Since stdlib is a public module, both /etc/facts.d and
/etc/puppetlabs/facts.d are scanned for static facts.
