On Fri, Feb 10, 2012 at 4:44 PM, Brock Noland <[email protected]> wrote:
> Hi, > > On Sat, Feb 11, 2012 at 12:34 AM, Karthik K <[email protected]> wrote: > > How about hadoop platform specific profiles ( and adding appropriate > > hadoop dependency ) , with classifiers as well: > > > > > http://www.sonatype.com/books/mvnref-book/reference/profiles-sect-tips-tricks.html > > . > > > > This should help generate maven artifacts with different classifiers > based > > on what people want to use , and should also update the artifact name > (with > > the classifier suffix) ? > > The only issue is that the maven lifecycle only compiles once, so you > have to hack around to get both jars in one tar.gz. It was recommended > we not do that here: > > https://issues.apache.org/jira/browse/MRUNIT-56 Curious, why we need to get both the jars (compiled against different hadoop platform versions) in a single tar.gz ? How about doing the mvn compilation / assembly/ deploy lifecycle once for each profile supported , to make sure they are indeed compiled against different versions, packaged and are separate ? And, people who want the appropriate versions can use the classifier to pull the relevant libraries or download the package itself ? -- Karthik.
