On 7/23/2013 10:44 AM, Jaroslaw Cwiklik wrote: > I dont recall as to why index=true is in the pom. I suspect it was pasted > in from some configuration > found on the web. It causes no harm as we are running with this setting > since day 1. But since since > it potentially may cause problems if our poms evolve I will remove it. Example harm: using a standard maven jar build to build a jar where you set additional class paths in the manifest - this fails because the part of the jar building which indexes the jars doesn't index the addition jars included in the manifest class-path entry.
-Marshall > > > > On Tue, Jul 23, 2013 at 10:28 AM, Marshall Schor <[email protected]> wrote: > >> Found another reason why setting index = true is not a good idea. >> >> If you have a Jar which specifies a classpath in its manifest, then you >> need >> special steps to insure that the index is built that includes the jars >> specified >> in the classpath. >> >> See >> http://closingbraces.net/2007/05/13/jarclasspathandindex/ >> >> But, it seems this is not needed for the case where the Jars are local, >> and the >> classes are all expected to be found. >> >> -Marshall >> >> >> On 6/19/2013 2:43 PM, Marshall Schor wrote: >>> The DUCC parent pom has a section for the maven-jar-plugin manifest >>> configuration which adds <index>true</index>. My reading of this is >> that it is >>> designed for collections of Jars - where the index is built in such a >> way as to >>> have all the entries for all the Jars, to enable applets to avoid >> downloading >>> Jars just to search their contents. (The 1st Jar has the index for all >> the Jars). >>> I don't think this usage in individual Jars helps anything, and just >> takes up >>> space... Other opinions? >>> >>> -Marshall >>> >>
