[ https://jira.codehaus.org/browse/MSHARED-396?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Ittai Zeidman closed MSHARED-396. --------------------------------- Resolution: Not A Bug I think it's not a bug in maven-archiver but in maven core. see my comment for more details > Created manifest contains versions of test scope > ------------------------------------------------ > > Key: MSHARED-396 > URL: https://jira.codehaus.org/browse/MSHARED-396 > Project: Maven Shared Components > Issue Type: Bug > Components: maven-archiver > Reporter: Ittai Zeidman > Attachments: dependency-bug.zip > > > Hi, > I'm not 100% sure this issue belongs here and not in maven-project itself but > seeing as this is where I'm having the issue I thought I'd start here. > When building the classpath for a manifest the chosen version of an artifact > is based on all scopes and not only on compile/runtime. > This is an issue in the following layout: > Main-Artifact > test-dependency-which-directly-depends-on-old-javassist (test) > javassist: 3.15.0-GA > compile-dependency-which-transitively-depends-on-new-javassist (compile) > dependency-which-directly-depends-on-new-javassist > javassist: 3.16.1-GA > I'm getting javassist 3.15.0-GA in my classpath and getting a method not > found during runtime when using methods from the newer version. > I would expect maven to do the "nearest path" resolution on the subset of > artifacts which actually compete on the same classpath (i.e. runtime). > I know that maven-archiver is just using project.getRuntimeClasspathElements > which uses project.getArtifacts which returns the old version but I think > maven-archiver can decide to use a different resolution mechanism. > I say this because we also use the maven-assembly-plugin to copy the > dependencies and that does the correct, IMO, resolution. > I've attached a multi-module project which demonstrates this. > Simply run jar:jar in the main-artifact module and take a look in the > manifest file of that jar [I'm using archiver via jar-plugin since that is my > use-case]. > I'm really willing to try and submit a patch or even just test-cases but I > wanted to know what you think of the issue first. -- This message was sent by Atlassian JIRA (v6.1.6#6162)