[jira] (MNG-3169) Dependency-Resolution Bug (Resolved too early / at wrong point)

["Jason van Zyl (JIRA)" <ji...@codehaus.org>]

     [ https://jira.codehaus.org/browse/MNG-3169?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jason van Zyl closed MNG-3169.
------------------------------

    Resolution: Cannot Reproduce

> Dependency-Resolution Bug (Resolved too early / at wrong point)
> ---------------------------------------------------------------
>
>                 Key: MNG-3169
>                 URL: https://jira.codehaus.org/browse/MNG-3169
>             Project: Maven 2 & 3
>          Issue Type: Bug
>          Components: Dependencies
>    Affects Versions: 2.0.6, 2.0.7
>            Reporter: Jörg Hohwiller
>             Fix For: Issues to be reviewed for 3.x
>
>
> I have a maven project with a 2 level deep module structure.
> The toplevel POM has two modules A and B that themselves have packaging pom and have several modules.
> The reactor build order is strict for A and B meaning that A and B are completely independent. So first A is build with all its modules and then B.
> If I do an 'mvn install' on the top-level project with a clean local repository the build fails because during the install of a module of A some 
> artifact of B is resolved, even though it is NOT referenced anywhere in a POM in A or below!
> If I do a mvn install in A it works fine.
> If I do a mvn install in B it works fine.
> After B is installed with all its modules, then the mvn install on the top-level project also works fine.
> The module in B that was missing first (B1) is required by another module of B (B2) that defines the dependency (on B1) twice, once regularly and a second time with classifier 'sources' since the sources are required for the GWT-Compiler to run.
> I tried to do a mvn -X install when the problem occurs but the log did NOT enlighten me anyhow. I only gave me the impression that the dependency resolution in maven is completely insane cycling around into the same things again and again and again and however ending without an infinity-loop in this strange bug. 
> What should '(selected for null)' say?
> Why doesnt the MavenProject declare a useable toString() method? I can not read stuff like 'org.apache.maven.project.MavenProject@fa682130'
> Maven in general is really cool. But if something goes wrong you are so very lost with it.



--
This message was sent by Atlassian JIRA
(v6.1.6#6162)