[
https://issues.apache.org/jira/browse/MNG-5761?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17277004#comment-17277004
]
Marc Carter commented on MNG-5761:
----------------------------------
Finally realised why I see so many poms with more {{<exclusions>}} than
{{<dependencies>}}. This is counter-intuitive and breaks basic encapsulation
principles.
To be clear, just in case it's lost in the above noise, my understanding here
is not that _all_ of {{project_A}}'s managed-dependencies should appear in
{{project_C}} (BOM-style). Only {{project_A}}'s _effective_
managed-dependencies should appear transitively - those that changed A's
_effective pom_.
I do see that this might be difficult to capture without publishing a generated
pom where those are instantiated as actual <dependencies> (sort of how
non-Maven tools publish into Maven repos by building a minimal "runtime pom").
> Dependency management is not transitive.
> ----------------------------------------
>
> Key: MNG-5761
> URL: https://issues.apache.org/jira/browse/MNG-5761
> Project: Maven
> Issue Type: Bug
> Components: Dependencies
> Affects Versions: 3.2.5
> Reporter: Jeff Schnitzer
> Priority: Critical
> Fix For: 4.0.x-candidate
>
> Attachments: MNG-5761.zip
>
>
> A detailed description of the issue is here:
> http://stackoverflow.com/questions/28312975/maven-dependencymanagement-version-ignored-in-transitive-dependencies
> The short of it is that maven appears to be using the wrong
> <dependencyManagement> version in a transitive dependency. There are two
> relevant <dependencyManagement> sections in the build, one pulled in by guice
> and one pulled in by gwizard-parent. These are the dependency paths from the
> top:
> gwizard-example -> gwizard-config -> gwizard-parent
> gwizard-example -> gwizard-config -> guice -> guice-parent
> gwizard-parent's dependencyManagement specifies guava 18
> guice-parent's dependencyManagement specifies guava 16
> Guava 16 is winning. This seems highly undesirable, and in fact it breaks our
> build. I would expect that in a version # fight, "closest to the top" should
> win.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
