Fixes #12390: Doubled "\" in GM call #1896
Conversation
| @@ -59,13 +59,13 @@ bundle server access_rules | |||
|
|
|||
| any:: | |||
| "/var/rudder/share/root/" | |||
| maproot => { host2ip("server.rudder.local"), string_downcase(escape("server.rudder.local")) }, | |||
| admit => { host2ip("server.rudder.local"), string_downcase(escape("server.rudder.local")) }, | |||
| maproot => { string_downcase(escape("server.rudder.local")) }, | |||
There was a problem hiding this comment.
That change should not be linked with the modification at hand, but I'm not sure why I need it. I'm wondering if we didn't change something elsewhere.
There was a problem hiding this comment.
strange, if not something in your PR then tests must failed but they passs ...
There was a problem hiding this comment.
I know, and I don't know what to think about it. @ncharles, do you have an idea?
There was a problem hiding this comment.
Found why, it's because we changed booleans in string template to string, which change conditionals behavior in stringtemplate
|
PR rebased |
fanf
force-pushed
the
bug_12390/doubled_in_gm_call
branch
from
fd3ac28
to
ff40e1c
Compare
| @@ -104,7 +99,7 @@ object VTypeConstraint { | |||
| sealed trait StringVType extends VTypeConstraint with VTypeWithRegex { | |||
| def regex: Option[RegexConstraint] | |||
|
|
|||
| override def getTypedValue(value:String, forField:String) : Box[Any] = regex match { | |||
There was a problem hiding this comment.
The Any was ... interesting ...
| @@ -192,12 +187,12 @@ final case object LinuxDerivedPasswordVType extends DerivedPasswordVType { overr | |||
|
|
|||
| case object BooleanVType extends VTypeConstraint { | |||
| override val name = "boolean" | |||
| override def getTypedValue(value:String, forField:String) : Box[Any] = { | |||
| try { | |||
| override def getFormatedValidated(value:String, forField:String, escapeString: String => String) : Box[String] = { | |||
There was a problem hiding this comment.
This is the cause of the issue in templates, the value was previous a boolean, not a string, then when used in conditional in string template, if considers truthy anything that has a value or check the boolean value if it's a boolean:
IF actions test the presence or absence of an attribute unless the object is a Boolean/bool, in which case it tests the attribute for true/false. The only operator allowed is "not" and means either "not present" or "not true". For example, "$if(!member)$...$endif$".
|
PR rebased |
fanf
force-pushed
the
bug_12390/doubled_in_gm_call
branch
from
ff40e1c
to
6097ea2
Compare
|
This PR is not mergeable to upper versions. |
|
OK, merging this PR |
| @@ -221,53 +221,22 @@ final case class Cf3PolicyDraft( | |||
| */ | |||
| case class ParameterEntry( | |||
| parameterName : String | |||
| , parameterValue: String | |||
| , escapedValue : String | |||
There was a problem hiding this comment.
if i'm correct, this is used for the global parameters - here we are droping the escaping of these parametres, causing https://www.rudder-project.org/redmine/issues/12674
https://www.rudder-project.org/redmine/issues/12390
This is an API breaking change for Windows DSC plugin.
So, the main problem was that we escaped value set in Directive only for a CFEngine Agent.
The pull request change two majors things.
First, it changes where the escape logic is in the code:
AgentRegisteris now a standalone class, with utility method to find what agent matches anAgentType/osDetailsand does things with it.Secondly, it changes the types of values in string template. Until now, we were allowing values of type
Any. But we don't actually use anything but string, and I can't apply string escape logic onAny. So now, we only deal with strings (which will certainly allow to simplify things in the long term - because yes,Stringis better thanAny, especially when we don't use other things).