[MSOURCES-115] aggregate-sources yields different results based on reactor

**[Raymond DeCampo](https://issues.apache.org/jira/secure/ViewProfile.jspa?name=raydecampo)** opened **[MSOURCES-115](https://issues.apache.org/jira/browse/MSOURCES-115?redirect=false)** and commented

The main idea here is that depending on whether a particular project is built from the parent project or on its own, the sources jar generated by aggregate-sources may be different.  IMO this attacks the repeat-ability of the build.  The root cause is that aggregate-sources pulls in all the sources in the reactor.

Consider a three layer structure.  One parent project at the top level, say foo-parent.  Then two child modules at the next level, say foo-client and foo-server.  Then some number of children of foo-client and foo-server where the actual Java code lives.

The poms for foo-server and foo-client have aggregate-sources set up.  If a build is executed directly on foo-client for example, then the resulting source foo-client-sources-*.jar will have the source from the children of foo-client.

If a build is executed from foo-parent, both foo-client and foo-server are included in the reactor.  Now the resulting foo-client-sources-*.jar includes the source code from foo-server's children as well as its own.

I would expect that the aggregate-sources target would only include the source of the children of the current project instead of the source from any project in the reactor.  I would also expect that the build should result identical artifacts whether or not it is invoked from the parent project or not.

---

**Affects:** 3.0.1