| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| |
|
|
|
|
|
| |
This is Luke's patch plus a change to fix a test that it broke. It creates
a new sub-scope off the top scope, called "settings" and adds each of the
environment's settings to it as variables, thus satisfying the ticket while
taking us one step further from being able to implement futures. *sigh*
|
| |
|
|
|
|
|
|
| |
RDoc's parser produces errors on this sort of statement:
def (variable).method
This patch wraps our occurances of those definitions with comments that
suspend RDoc parsing.
|
| |
|
|
|
| |
This patch (via Nicholas Veeser) adds more debugging when SSL cert
verification fails.
|
| |
|
|
|
| |
The logic which iterates over the searchpath in reverse does not translate the name.
Therefore file overrides in :master or :agent are not picked up.
|
| |
|
|
|
|
| |
[4123] [4124] - combined unit test for both fixes since they share some common code.
proper unit tests to verify features for both patches.
|
| |
|
|
|
|
| |
Added hasstatus => true as attribute for new provider instance in init.
redhat checks the hasstatus in the provider to determine service status.
|
| |
|
|
|
|
|
|
| |
filenames to exclude when processing defpath.
also updated redhat provider to pass in a list of services to ignore.
didnt need to munch exclude to an array, include? is safe to call on strings
|
| |
|
|
|
|
| |
The initial commit changed the name of a method (close -> close_all) and
changed the way the array log destination worked before we saw that the
unit tests were using it differently.
|
| |
|
|
|
| |
The log will now queue any log messages created when there is no
destination, and will flush the queue when a destination is added.
|
| |
|
|
|
|
|
|
| |
This patch introduces Type#retrieve_resource as a wrapper for
Type#resource, to coerce the return value from legacy types from Hash to
Resource.
Signed-off-by: Jesse Wolfe <jes5199@gmail.com>
|
| |
|
|
| |
Reworked it to use the new key_attributes instead.
|
| |
|
|
| |
This patch is from Joe McDonagh <joseph.e.mcdonagh@gmail.com>
|
| |
|
|
|
| |
Changed the error message when searching for an Application constant
which is undefined.
|
| |
|
|
|
|
| |
By asking the environment for known resources instead of creating a type
collection ourselves, we avoid accidentally creating two
Resource::TypeCollection objects.
|
| |
|
|
|
|
| |
The hook for storeconfig will now require 'puppet/rails' if the setting
is set to true. It was previously being indirectly required via
parser/interpreter, which was removed.
|
| | |
|
| |
|
|
|
| |
We had a hardcoded assumption that the version number would always start with
a zero, and thus were failing to recognise 2.6.0 as greater than 0.25.x
|
| |
|
|
| |
Somehow one use of JSON escaped the global find and replace of PSON --> JSON.
|
| | |
|
| |
|
|
|
| |
Our settings were slow because I was querying Application objects for
their run_mode repetitively
|
| |
|
|
|
|
| |
Added a Puppet::Util::Settings layer called :mutable_defaults to emulate
the interaction between Puppet::Application and defaults.rb that was
getting thwarted by rspec.
|
| |
|
|
|
|
|
| |
assert their run_mode
Also cleanup of an unecessary puts line, make master tests run when
on their own, and moving a require to a more usual spot.
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
| |
Previously, any failed call to Puppet::Application.find would result
in an error being raised by const_get, resulting in a messy crash with
a stack trace. Now that error is handled, and the application will
print a message and exit.
|
| |
|
|
|
|
| |
Notify was returning a nil current value, rather than :absent,
which caused puppet to think the message had been changed from
blank to its new value, rather than defined as that value.
|
| |
|
|
|
| |
The describe and resource (ralsh) applications required puppet, creating a
loop & thus crashing them.
|
| |
|
|
|
|
|
| |
RunMode is only created once instead of every time it's called
Got lots of unpredictable test failures, presumably because a new
RunMode was being created every time we accessed the RunMode.
|
| | |
|
| |
|
|
|
| |
Along the way this fixes an issue with 2.6 alpha that sections of the
puppet config file were getting ignored.
|
| |
|
|
|
|
|
|
| |
Mode is a terribly overused word. Files use it, puppetdoc uses it, and
certs use it, and those are just the places that I happened to
stumble upon. It makes reading code very confusing and finding things
in code difficult. I know namespacing allows us to reuse words for
method and variable names, but that doesn't mean we should.
|
| |
|
|
|
|
| |
Per Luke's replacement of :check with :audit, and deprecation of :check,
I've replaced all of our internal uses of :check with :audit. Importantly,
this silence the deprecation warnings during regular usage from eg. ralsh.
|
| | |
|
| | |
|
| |
|
|
| |
I'm also going to update the wiki to mention sqlite is a test dependency
|
| |
|
|
| |
Signed-off-by: Luke Kanies <luke@puppetlabs.com>
|
| |
|
|
|
|
|
|
| |
We were looking only to the class hierarchies when trying to find an
ancestor to the current type. Thus we were never trying to climb up
the hierarchy of nodes when evaluating nodes.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| |
|
|
|
|
|
|
| |
During the refactoring of AST hostclass/node to non AST objects, we lost
the fact that the main class already comes with a scope (ie the top one),
so when we evaluate its code we shouldn't create a subscope for it.
Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
|
| |
|
|
|
|
|
|
| |
There was a subtle 1.8.7 dependence in the composite key / namevar patch;
Nick discovered that our assumtion that hashes could be used as hash keys
does not hold in earlier bersions of ruby. This patch replaces the hash
valued uniqueness_key with an array of the values of the (ordered) key
attributes.
|
| |
|
|
|
|
| |
Merging some of the branches creates a require-loop; this commit breaks it by
removing the most recently added link (which I believe to have been unneeded /
unrelated to the patch in which it snuck in).
|
| |
|
|
| |
There was a comma missing from a raise, which caused a warning message on each run.
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
and update the specs that were testing it.
|
| |
|
|
|
| |
The value method was failing to return any value in the case where
the value was false.
|
| | |
|
| | |
|
| | |
|
