summaryrefslogtreecommitdiffstats
path: root/spec/unit
Commit message (Collapse)AuthorAgeFilesLines
...
* Not trying to load files that get removed in pluginsyncingLuke Kanies2009-05-261-0/+12
| | | | | | | Previously any changed file got loaded; now we only try to load files that are still present. Signed-off-by: Luke Kanies <luke@madstop.com>
* Fixing #2288 - fixing the tests broken by my attr_ttl codeLuke Kanies2009-05-221-0/+9
| | | | Signed-off-by: Luke Kanies <luke@madstop.com>
* Fix for #2234: test fails with old Rack versionChristian Hofstaedtler2009-05-203-4/+4
|
* Fixing #2273 - file purging works more intuitivelyLuke Kanies2009-05-202-17/+31
| | | | | | | | | | | | | | It now correctly purges files whether we're recursing locally or remotely. *Please* test various scenarios you can think of with this. I've tested: * Local recursion with no remote source * Remote recursion with a source * Recursion with an extra locally managed file Signed-off-by: Luke Kanies <luke@madstop.com>
* Caching whether named autoloaded files are missingLuke Kanies2009-05-202-1/+31
| | | | | | | | This is the big win, because it causes us to just skip the whole loading infrastructure, including skipping looking through the modulepath. Signed-off-by: Luke Kanies <luke@madstop.com>
* Adding caching of file metadata to the autoloaderLuke Kanies2009-05-202-2/+141
| | | | | | | | | | | The cache isn't actually used yet - this just adds all of the plumbing. It was found that stat'ing files that didn't exist could take up to 85% of a run, so this is progress toward getting rid of those stats. Signed-off-by: Luke Kanies <luke@madstop.com>
* Adding modulepath caching to the AutoloaderLuke Kanies2009-05-201-2/+48
| | | | | | | | | There's more caching to add, but this simplifies the interface to the list of paths and then caches that list so we aren't constantly searching the filesystem. Signed-off-by: Luke Kanies <luke@madstop.com>
* Adding caching to the Environment classLuke Kanies2009-05-201-0/+16
| | | | | | | | | | | Caching the module path (because we check which directories exist, and this method can get called often), and the complete list of modules. The cache ttl uses the filetimeout, which defaults to 15 seconds. Signed-off-by: Luke Kanies <luke@madstop.com>
* Adding TTL support to attribute cachingLuke Kanies2009-05-201-1/+53
| | | | | | | | | Previously you had to have an Expirer, but now you can declare a TTL for a cached attribute and it will be expired automatically when the cached value is older than the ttl. Signed-off-by: Luke Kanies <luke@madstop.com>
* Deprecating factsync - pluginsync should be used insteadLuke Kanies2009-05-201-0/+8
| | | | Signed-off-by: Luke Kanies <luke@madstop.com>
* Protecting Stomp client against internal failuresLuke Kanies2009-05-181-0/+5
| | | | | | | | Apparently the stomp client is really unhelpful with failures; this attempts to provide at least a bit more information. Signed-off-by: Luke Kanies <luke@madstop.com>
* Refactoring the stomp client and tests a bitLuke Kanies2009-05-181-32/+87
| | | | | | | | | | | | | The main goal of this refactor is to tell the client to be resilient to failures (configured at initialization time), and to send all messages as persistent messages (configured for each message). In the process, the client now parses the queue source URI and handles each argument separately. The tests are more thorough, also. Signed-off-by: Luke Kanies <luke@madstop.com>
* Relying on threads rather than sleeping for puppetqdLuke Kanies2009-05-181-2/+3
| | | | | | | | We previously manually slept, but this uses the queue client to handle keeping the process running, by just joining all running threads. Signed-off-by: Luke Kanies <luke@madstop.com>
* Fixing #2250 - Missing templates throw a helpful errorLuke Kanies2009-05-182-15/+31
| | | | | | | | | | | This changes the behaviour of template searching a bit - we previously usually returned a file name, whether the template existed or not. Now we only return a path if it exists. Refactoring a few of the the tests for TemplateWrapper, also. Signed-off-by: Luke Kanies <luke@madstop.com>
* Fixing #2273 - CA location is set correctly in puppetcaLuke Kanies2009-05-171-0/+6
| | | | | | It's set to 'only' instead of 'local'. Signed-off-by: Luke Kanies <luke@madstop.com>
* Modules now can find their own pathsLuke Kanies2009-05-154-82/+170
| | | | | | | | | | | | Previously, when you created a module you had to specify the path. Now Module instances can use the module path to look up their paths, and there are methods for determining whether the module is present (if the path is present). Also cleaned up the methods for figuring out what's in the module (plugins, etc.). Signed-off-by: Luke Kanies <luke@madstop.com>
* Moving file-searching code out of Puppet::ModuleLuke Kanies2009-05-153-181/+196
| | | | | | | | | | The Module class had a bunch of code for finding manifests and templates even when not in a module, and it complicated the class unnecessarily. This moves that code to a new, hackish-but-sufficient module for just that purpose. Signed-off-by: Luke Kanies <luke@madstop.com>
* Fixing #2221 - pluginsignore should work againLuke Kanies2009-05-151-1/+7
| | | | | | | | We weren't splitting on whitespace, which is necessary since the settings don't support arrays but files expect them. Signed-off-by: Luke Kanies <luke@madstop.com>
* Confine stomp tests to Stomp enabled systemsBrice Figureau2009-05-151-0/+4
| | | | Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Fix bug #2124 - ssh_authorized_key always changes target if target is not ↵Francois Deppierraz2009-05-151-1/+8
| | | | defined
* Fix #1409 once again, including testFrancois Deppierraz2009-05-061-0/+9
| | | | | | | "user doesn't exit" error appeared once again after the changes which were applied in order to fix #2004. Validation must only check attributes presence, not their value.
* Fixing #2200 - puppetqd expects Daemon to be a classLuke Kanies2009-05-041-0/+185
| | | | | | | | | | | I *swear* I wrote tests for the daemon, but I can't find them in any of my branches so I rewrote them. In the course of writing them, I also fixed the usage of Daemon. Signed-off-by: Luke Kanies <luke@madstop.com>
* Removing unneeded test stubsLuke Kanies2009-05-031-8/+0
| | | | Signed-off-by: Luke Kanies <luke@madstop.com>
* Fixing #2195 - the Server class handles bindaddressLuke Kanies2009-05-031-25/+35
| | | | | | | The Server class has all of the logic now, instead of doing weird things in the defaults. Signed-off-by: Luke Kanies <luke@madstop.com>
* puppetmasterd can now run as a standard Rack application (config.ru-style)Christian Hofstaedtler2009-05-021-0/+41
|
* Add XMLRPC compatibility for RackChristian Hofstaedtler2009-05-022-0/+188
|
* Puppet as a Rack applicationChristian Hofstaedtler2009-05-022-0/+265
| | | | | | | This lays the ground: a wrapper for the REST handler, and an application confirming to the Rack standard. Also includes a base class for Rack handlers, as RackREST will not stay the only one, and there needs to be a central place where client authentication data can be checked.
* Add an unmunge capability to type parameters and propertiesBrice Figureau2009-05-022-0/+26
| | | | | | | | | | | | | Unmunge is the reverse of munge. While munge allows the type to return a different parameter value or properties should than the one it was created with, unmunge does the reverse. It can be used for instance to store a value in a different representation but still be able to return genuine value to the outside world. Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Fix #2218 - Ruby YAML bug prevents reloading catalog in puppetdBrice Figureau2009-05-022-4/+21
| | | | | | | | | | | | | | | | | | | Because of ruby bug: http://rubyforge.org/tracker/?group_id=426&atid=1698&func=detail&aid=8886 and http://redmine.ruby-lang.org/issues/show/1331 YAML dump of hashes using ruby objects as keys is incorrect leading to an error when deserializing the YAML in puppetd. The error is easy to correct by a post-process fix-up of the generated YAML, which transforms: &id004 !ruby/object:Puppet::Relationship ? to the correct: ? &id004 !ruby/object:Puppet::Relationship Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Fix #2207 - type was doing its own tag management leading to subtile bugsBrice Figureau2009-04-291-0/+26
| | | | | | | | | This patch moves Type to use Puppet::Util::Tagging as the other part of Puppet. This brings uniformity and consistency in the way the tags are used and/or compared to each other. Type was storing tags in Symbol format, which produced #2207. Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Fixing #2183 - checksum buffer size is now 4096bLuke Kanies2009-04-241-1/+1
| | | | Signed-off-by: Luke Kanies <luke@madstop.com>
* Fixing #2187 - Puppet::Resource is expected by Rails supportLuke Kanies2009-04-242-0/+91
| | | | | | | | | | | | | We previously used and expected Puppet::Parser::Resource instances, but 0.25 converts them all to Puppet::Resource instances before they're passed out of the compiler, so the Rails integration had to be changed to expect that. There's still some muddling, because the rails resources only generate parser resources, but that works for now because that's what we expect when collecting resources. Signed-off-by: Luke Kanies <luke@madstop.com>
* Adding an 'Exported' attribute to Puppet::ResourceLuke Kanies2009-04-241-0/+7
| | | | | | This is required for Rails support. Signed-off-by: Luke Kanies <luke@madstop.com>
* Making sure the cert name is searched firstLuke Kanies2009-04-241-9/+12
| | | | | | | | | | | The cert name should be searched first in default circumstances, even if it disagrees with the hostname. Brice's change to the way catalogs are searched for didn't quite work when the hostname and certname didn't agree *and* the certname was fully qualified. Signed-off-by: Luke Kanies <luke@madstop.com>
* Removing deprecated concurrency setting usage in railsLuke Kanies2009-04-241-7/+0
| | | | Signed-off-by: Luke Kanies <luke@madstop.com>
* Always making sure graph edges appear firstLuke Kanies2009-04-241-0/+7
| | | | | | | If we don't do this, there's a chance we'll get hit by the ruby yaml bug again. Signed-off-by: Luke Kanies <luke@madstop.com>
* Modifying the Settings#handlearg prototypeLuke Kanies2009-04-251-0/+14
| | | | | | | | I'd made the argument no longer optional because I thought the method was rarely used, but it's used in puppetd a good bit. Signed-off-by: Luke Kanies <luke@madstop.com>
* Fix configurer to retrieve catalog with client certnameBrice Figureau2009-04-231-3/+3
| | | | | | | | | | | | | | | | | Rationale: Before this change, the catalog was retrived with this uri: /catalog/hostname On the server side, the corresponding node was found by using the request node, then finding if this node also match hostname (which it does of course). But it is not possible to have an ACL matching the hostname part of the uri, because it: * it would be compared to the node name (certname), which obviously is not the same * it is not possible to create a dynamic allow/deny rule on a non-fqdn Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Unify auth/unauthenticated request authorization systemBrice Figureau2009-04-232-52/+48
| | | | | | | | | | | | | Before this change, unauthenticated REST requests where inconditionnaly allowed, as long as they were to the certificate terminus. This could be a security hole, so now the REST requests, authenticated or unauthenticated are all submitted to the REST authorization layer. The default authorizations now contains directives to allow unauthenticated requests to the various certificate terminus to allow new hosts. The conf/auth.conf file has been modified to match such defaults. Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Allow REST auth system to restrict an ACL to authenticated or ↵Brice Figureau2009-04-232-5/+86
| | | | | | | | | | | | | unauthenticated request Introduces a new auth.conf directive (auth or authenticated) which takes an argument (on,yes/off,no/all,any). This can be used to restrict an ACL to only some state of authentication of a REST request, or any. If no auth directive is given, the ACL will only trigger for authenticated requests. Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Fill REST request node with reverse lookup of IP addressBrice Figureau2009-04-233-4/+30
| | | | Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Refactor rest authorization to raise exceptions deeperBrice Figureau2009-04-235-75/+113
| | | | | | | | | | | | | The idea is to raise an AuthorizationException at the same place we check the authorization instead of in an upper level to be able to spot where the authorization took place in the exception backtrace. Moreover, this changes also makes Rights::allowed? to return the matching acl so that the upper layer can have a chance to report which ACL resulted in the match. Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Add environment support in the REST authorization layerBrice Figureau2009-04-233-6/+53
| | | | | | | | | With the help of the new auth.conf directive 'environment', any ACL can now be restricted to a specific environment. Omission of the directive means that the ACL will apply to all the defined environment. Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Fix #1875 - Add a REST authorization systemBrice Figureau2009-04-233-0/+207
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | This patch introduces a new configuration file (and configuration setting to set it). Each REST request is checked against this configuration file, and is either allowed or denied. The configuration file has the following format: path /uripath method <methods> allow <ip> or <name> deny <ip> or <name> or path ~ <regex> method <methods> allow <ip> or <name> deny <ip> or <name> where regex is a ruby regex. This last syntax allows deny/allow interpolation from the regex captures: path ~ /files[^/]+/files/([^/]+)/([^/])/ method find allow $2.$1 If you arrange your files/ directory to have files in 'domain.com/host/', then only the referenced host will be able to access their files, other hosts will be denied. For instance: files/reductivelabs.com/dns/... files/reductivelabs.com/www/... then only files in dns can be accessible by dns.reductivelabs.com and so on... If the auth.conf file doesn't exist puppet uses sane defaults that allows clients to check-in and ask for their configurations... Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Enhance authconfig format to support uri paths and regexBrice Figureau2009-04-232-21/+453
| | | | | | | | | | | | | | | | | | | | | This patch introduces a new set of directive to the authconfig parser/file format: path /uripath or patch ~ <regex> This directive declares a new kind of ACL based on the uri path. method save, find This directive which is to be used under path directive restricts a path ACL to only some REST verbs. The ACL path system matches on path prefix possible, or on regex matches (first match wins). If no path are matching, then the authorization is not allowed. The same if no ACL matches for the given REST verb. The old namespace right matching still works as usual. Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Add RSpec unit tests for network rightsBrice Figureau2009-04-231-0/+54
| | | | Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Add RSpec unit tests for authconfigBrice Figureau2009-04-231-0/+173
| | | | Signed-off-by: Brice Figureau <brice-puppet@daysofwonder.com>
* Adding support for specifying a preferred serialization formatLuke Kanies2009-04-221-0/+11
| | | | | | | | | | This isn't that great, but at least it provides basic tuning of the format. Also removing the catalog_format default, since it's no longer used. Signed-off-by: Luke Kanies <luke@madstop.com>
* Fixing the FormatHandler test to use symbols for format namesLuke Kanies2009-04-221-22/+25
| | | | | | | Formats guarantee that symbols are used, so it makes sense for the tests to do so. Signed-off-by: Luke Kanies <luke@madstop.com>
* Fixing #2149 - Facts are passed as part of the catalog requestLuke Kanies2009-04-224-22/+94
| | | | | | | | | | | | | | | This removes the requirement of shared fact caching on the servers, since the server responding to the catalog request will receive the facts as part of the request. The facts are serialized as a parameter to the request, rather than each being set as a separate request parameter. This hard-codes yaml as the serialization format for the facts, because I couldn't get marshal to work and it's just not as big a deal for such a small amount of data. Signed-off-by: Luke Kanies <luke@madstop.com>
generated by cgit (git 2.34.1) at 2025-09-28 23:11:13 +0000