Some modules or module versions don't have a metadata.json file, but we
might still want to use the load_module_metadata function on them. The
lack of a file can still give us important information. For example, it
might tell us that the version of the module installed is "very old"
even if we can't read the version number directly. This patch adds a
parameter to let the user specify if an empty file is acceptable. To
preserve backwards compatibility it does not change the current default
behavior, which is to raise an error if metadata.json does not exist.
This is to improve speed on Facter resolution of service_provider fact
that was just introduced in PR# 506. The improvements go from 280ms
resolution time approx. down to 2ms resolution time approx. by adding
requires statements.
This adds a package_provider fact for situations where we need to be
able to know the client's package provider in a simple way. Situations
such as: package { 'name': install_options => [] } As those tend to be
package provider specific options.
Maintain the old behavior in the case where the optional second
parameter isn't passed. Also, adding arity is backwards incompatible since
stdlib still supports 2.7, so remove that.
The data type system is very hard to understand. Many people don't
understand why
type_of([1,2,3]) == Array
will fail, but
type_of([1,2,3]) <= Array
passes. This does a simpler validation that doesn't rely on explicit
data types. Instead, use
$foo = [1,2,3]
if $foo.is_a(Array) {
notify { 'This is an array': }
}
This is based on code by Ben Ford <ben.ford@puppetlabs.com>.
* Added acceptance tests
* Added dispatch
* Improved unit tests
* Added docs to README
* previously the str2bool function did not accept 'TRUE' as a bool
type. This causes the function to now accept TRUE, FALSE strings
as a boolean type in order to be converted to a proper boolean.
* This would also cause Y,N, YES, NO to be accepted as boolean types
as well.
Add spec tests to test the new functionality:
*Case for 3 arrays.
*Case for 4 arrays.
Modify README to note new functionality.
This is for issue MODULE-2456, follow the precedent of MODULE-444.
This change allows union to be much more useful, unioning many arrays
in one line rather than in n lines. Additionally, as this is only added
functionality, and does not affect the 2 array case that all modules
currently using array are using, it should not affect any existing
modules utilizing union.
This is now useful, for example, for merging many arrays of resources
(eg: packages.) to generate just one list with no duplicates, to avoid
duplicate resource declarations.
This returns the default provider Puppet will choose to manage services
on this system by instantiating a dummy service resource type and
returning the provider chosen.
Co-Authored-By: Simon Fraser University <asa188@sfu.ca>
Split the `destroy` method of the file_type::ruby provider into two
private methods: `handle_destroy_line` which is the same as the previous
`destroy` method, and `handle_destroy_with_match` which will destroy any
line which matches the `match` parameter, raising an error if multiple
lines match and the `multiple` parameter is not `true`. This new
behavior is only used if the new boolean parameter `match_for_absence`
is `true` (it defaults to `false`).
Puppet's boolean parameter type is only available in Puppet 3.3 and
higher, so change file_type's new "replace" parameter to a regular
parameter with true and false as possible values. This matches the
existing "multiple" parameter.
This function loads the metadata.json into a puppet variable. This enables a number of neat things such as:
* Which version of the module am I using? 2.x? 3.x?
* Which author of the module am I using? puppetlabs? example42?
The documentation example shows an incorrect response when using the
function, this PR corrects the example to agree with what the function
actually does.
facter (2.x) only provides facts without interface suffix for
* ipaddress
* netmask
'macaddress' and 'network' facts will always have the related interface
name appended. in turns lookupvar throws errors when strict_variables is
enabled.
The time() function takes an argument of a timezone, and always returns
time in epoch format. The epoch format is the number of seconds that
have elapsed since January 1, 1970 (midnight UTC/GMT), not counting leap
seconds. This means that it is universally the same regardless of
timezones.
I don't know what the timezone argument is supposed to do, and it is not
documented. So lets just make 1.8.7 work like > 1.8.7
Previously, the random number generator was seeded with the array or
string to be rotated in addition to any values specifically provided for
seeding. This behavior is potentially insecure in that it allows an
attacker who can modify the source data to choose the post-shuffle
order.
Tests in the new style produces the following documentation output:
abs
should not eq nil
should run abs() and raise an Puppet::ParseError
should run abs(-34) and return 34
should run abs("-34") and return 34
should run abs(34) and return 34
should run abs("34") and return 34
Without this, file_line resource without the `match` parameter but with the `after` param will throw an error if there are multiple matches for the after expression. This patch creates the handling for the `multiple` parameter in handle_create_with_after. This allows you to add a line after the `after` expression if it appears at multiple points in a file.
Updated reference to `file_line` in the README to reflect that the multiple parameter can be set when using `after` and/or `match` as the matching regex.
Without this, the global seed is reseeded on every use
of fqdn_rotate, which is a waste. Older rubies might even use a
time-base seed which adversly impacts the quality of the RNG.
As mentioned in #270, private is a reserved keyword in the future parser
which is to be released with Puppet 4. As it stands, this function is
not useable with the future parser so it needs to renamed.
This is a breaking change.
The previous commit to uriescape() changed the implementation to use the ruby default escape list for URI.escape(), but did not change the call triggered when uriescape() was called on an array, triggering ruby errors.
As per puppetlabs/puppet@292233c, this leaves the global seed in a
deterministic state, which is bad. Puppet::Util.deterministic_rand()
exists to avoid running into this issue, but is only present starting in
Puppet 3.2.0.
The `type()` function will cease to work on the new parser because 'type'
is a reserved keyword. The `type3x()` function may be used to continue
similar functionality, but will be deprecated in favor of the built-in
typing system.
The `type_of()` function has been included to introspect types in the
new parser.
The `match` attribute was validated to match `line`, except that in many
cases (even the example given in the docs) a user would want to match a
line entirely different from the new line.
See comments on the original commit
a06c0d8115
and ask
https://ask.puppetlabs.com/question/14366/file_line-resource-match-problems/
for further examples of confusion.
