While looking into SOLR-4852 and testing every conceivable lib permutation, I ran across a second problem, I'd like to know if it should be considered a bug.
https://issues.apache.org/jira/browse/SOLR-4852?focusedCommentId=13667025#comment-13667025 What I was trying to do here was split my required jars between ${solr.solr.home}/lib and ${solr.solr.home}/foo ... the former directory is automatically used for libraries, the latter was added by sharedLib="foo" in my solr.xml. Should this be a valid configuration? If not, perhaps we need to stop automatically including ${solr.solr.home}/lib. I run into the same problem (unable to find the ICUTokenizer class) whenever I split my jars, even though the icu analysis jar was not the jar that I moved. When I first tried it, I moved the icu4j jar, but it also has the exact same problem when I move the mysql jar, which has nothing at all to do with ICU. Here's a Solr log (on an unpatched branch_4x) from when I moved the mysql jar from lib to foo. You can see the jars that get loaded, so this should not be happening: http://apaste.info/6aK5 If all the jars are in either lib or foo, everything works. Is this behavior a bug? I am starting to think that this problem and the original SOLR-4852 issue are actually the same problem, and that it may not be a duplicate jar problem, but rather something specific and subtle with the ICU analysis components that happens when the classloader is replaced. Thanks, Shawn --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
