- Updated to add 4.0 to travis-ci configs
- Add .sync.yml for modulesync usage

(FM-2752) Add/update travis with modulesync

(maint) Add beaker-puppet_install_helper

 - When running against current versions of Puppet that use Ruby 2.x,
   this causes issues installing the bundle.  Guard for backward compat
   with Ruby 1.9.x / Puppet 3.7 x86

 - 'simplecov' requires the 'json' gem, which requires native build
   tools to be installed.  Current Windows CI spec boxes intentionally
   do not install these build tools.  Since 'simplecov' is unused, it
   can be removed.

…c-box-compatibility

Maint/master/gemfile windows spec box compatibility

(FM-2978) Update beaker setting log_level => debug

The version of Puppet/Facter/Hiera will determine the versions of dependencies,
namely the Windows gem dependencies, so relax the versioning at bit to allow for
a wider range of gems to be pulled in across many versions of Puppet that may
be supported by the module.

We fully support specifying Puppet, Facter, and Hiera versions as environment
variables, and Bundler should "do the right thing" when resolving dependencies.
This comes with the following considerations:

* Explicitly require Windows gems when we are using a git or file location for
  Puppet, or are requiring a Puppet version less than `3.5.0` (the first version
  of Puppet that had a Windows platform-specific gem).
* The resolution for versions less than `3.5.0` is very naive and will only
  match on the version being specified as a version (e.g. '3.4.3') and not a
  version spec (e.g. `~> 3.4`).
* These versions may be file:// for Puppet, Facter, OR Hiera, or All of them to
  support the scenario of making changes to Puppet in concert with local module
  changes.

Other things to note:

* Bundler behaves better when omitting a dependency that is not necessary to
  explicitly include.  For instance, adding `Hiera >= 0`, when it was not
  specified in an environment variable, resulted in some rather surprising
  Bundler behavior.  Namely it resolved an odd Hiera version, which caused a
  cascading effect of incorrect dependency resolution for other gems
  (namely win32-security) including Puppet.
* Right now, TravisCI exhibits some weird behavior with installing Facter `1.7.6`
  for `Puppet ~> 3.0` (presently `3.8.1`) because it appears to be looking at
  the universal-darwin Facter gem.  Beaker transitively requires
  `CFPropertyList ~> 2.2`, which installs `2.3.1`.  This later conflicts with
  the `CFPropertyList ~> 2.2.6` requirement that newer Facter versions have.
  Bundler works backward to find a Facter gem that doesn't use CFPropertyList,
  and ends up resolving `1.7.6`.
* If you need to better understand Bundlers resolution mechanism, set the
  `DEBUG_RESOLVER` environment variable to `1`.

…-relax-compatibility

(maint) Relax gem dependencies for Windows

 - Setup a matrix with:
    - Ruby 2.1 and Puppet 4.x (latest) and
    - Ruby 2.0 x64 / Ruby 1.9.3 x86 + Puppet 3.x (latest)
    - Ruby 1.9.3 x86 + Puppet 3.3.2 (oldest supported)

(maint) add AppVeyor support

This ensures the following items are under moduleysnc:

.gitignore
.travis.yml
CONTRIBUTING.md
Gemfile
appveyor.yml
spec/spec.opts

Most importantly - this ensures that acl has up to date Gemfile, Travis CI and
AppVeyor configurations that allow it to be tested properly.

(MODULES-2207) Update Modulesync

 - As it turns out, the previous test for a well known value
   would fail under Windows 2003, because the value is SYSTEM
   instead of System.  We could adjust the case, but instead
   perform a match of a given key to an environment variable
   that should match it to rule out OS specific casing issues.

(maint) fix registry_value provider spec fail

Further restrict down the restrictions on windows dependencies when
using older Puppet versions, such that it works with Puppet 3.0.0 -
3.7.0. It does require a bit more logic in selecting gems but it allows
for testing older platforms without any additional known issues for
Puppet versions under 3.4.0.

There is an issue with rspec 2.14.0 - it doesn't have
`any_instance_recorder_for` which rspec-puppet uses. The fix is simply
to move the dependency to `~> 2.14.1` which does have
`any_instance_recorder_for`.

(MODULES-2207) Gem restrictions for Older puppet versions

Using PSH with versions lower than 0.9.0 results in issues
attempting to symlink on Windows. We'll take the latest
version as the dependency since it has the best fixes
available for cross platform testing.

(maint) puppetlabs_spec_helper ~>0.10.3

 - Beakers shell function always uses the default node, which is
   undesirable for installing stdlib
 - The x64? check in should_manage_values should be using the bitness
   of the node under test, not the default node

 - The Ruby implementation of write is buggy at :
  https://github.com/ruby/ruby/blob/v2_1_6/ext/win32/lib/win32/registry.rb#L727-L748

  When writing a REG_MULTI_SZ, the MSDN documentation for RegSetValueEx
  https://msdn.microsoft.com/en-us/library/windows/desktop/ms724923(v=vs.85).aspx
  states that for the lpData parameter that "With the REG_MULTI_SZ data
  type, the string must be terminated with two null characters."  It
  further states for the cbData parameter that "The size of the
  information pointed to by the lpData parameter, in bytes. If the data
  is of type REG_SZ, REG_EXPAND_SZ, or REG_MULTI_SZ, cbData must include
  the size of the terminating null character or characters."

  Looking at the implementation, we can see that the length for a
  REG_MULTI_SZ is calculated properly, but the second terminating NULL
  is never written to the string.

  The REG_SZ and REG_EXPAND_SZ handling is affected by the same issue as
  the length calculation takes into account a terminator that is not
  present.

  This can lead to intermittent memory corruption based on what's
  present in the last 2 bytes of a given buffer. There is no guarantee
  that this memory is zeroed out prior to use.

  Unfortunately, the only way to work around this issue is to do a
  reimplementation of write that terminates the array of bytes properly
  (with a double NULL terminator that is UTF16-LE).  Note that
  the end of the byte array should be 4 NULL bytes like [0, 0, 0, 0]

 - Add additional specs for writing and destroying all of the supported
   registry types.

 - Note: No support has been added for REG_DWORD_BIG_ENDIAN

…registry-write-corruption

(MODULES-1921) Ruby registry writes corrupt string

(FM-3089) Prepare 1.1.1 release