https://issues.apache.org/jira/browse/SOLR-5173
On Aug 18, 2013, at 8:43 PM, Steve Rowe <sar...@gmail.com> wrote: > bq. I thought that when Steve moved it from the test module to the core, he > handled it so that it would not go out in the dist. > > Mea culpa. > > @Chris Collins, I think you're talking about Maven dependencies, right? As a > workaround, you can exclude dependencies you don't need, including > hadoop-hdfs, hadoop-auth, and hadoop-annotations - this will also exclude the > indirect jetty 6 dependency/ies. hadoop-common is a compile-time dependency, > though, so I'm not sure if it's safe to exclude. > > The problems, as far as I can tell, are: > > 1) The ivy configuration puts three test-only dependencies (hadoop-hdfs, > hadoo-auth, and hadoop-annotations) in solr/core/lib/, rather than where they > belong, in solr/core/test-lib/. (hadoop-common is required for solr-core > compilation to succeed.) > > 2) The Maven configuration makes the equivalent mistake in marking these > test-only hadoop dependencies as compile-scope rather than test-scope > dependencies. > > 3) The Solr .war, which packages everything under solr/core/lib/, includes > these three test-only hadoop dependencies (though it does not include any > jetty 6 jars). > > 4) The license files for jetty and jetty-util v6.1.26, but not the jar files > corresponding to them, are included in the Solr distribution. > > I have working (tests pass) local Ant and Maven configurations that treat the > three hadoop test-only dependencies properly; as result, the .war will no > longer contain them - this will cover problems #1-3 above. > > I think we can just remove the jetty and jetty-util 6.1.26 license files from > solr/licenses/, since we don't ship those jars. > > I'll open an issue. > > Steve > > > > On Sun, Aug 18, 2013 at 1:58 PM, Mark Miller <markrmil...@gmail.com> wrote: > > On Aug 17, 2013, at 9:01 AM, Robert Muir <rcm...@gmail.com> wrote: > > > I think this is only a "test dependency" ? > > Right - it's only for the hdfs 'test' setup. I thought that when Steve moved > it from the test module to the core, he handled it so that it would not go > out in the dist. > > - mark > >
