: return IndexReader.open(FSDirectory.open(new : File(h.getCore().getIndexDir())), true); : } : : *org.apache.lucene.index.IndexNotFoundException: no segments* file found in
I haven't dug into it, but i seem to recall that there were some changes a while back to the TestHarness and solr base test classes to start using Randomized Directory impls that help with verifying the code on more varieties of platforms. I suspect that it's also using a RAM Based directories, and that nothing is ever actually getting written to the getIndexDir() directory. IN general, now that multiple types of IndexReader and DirectoryProvider factories are supported, you can't (in general) assume that SolrCore.getIndexDir() will actually exist (let alone be used) As a side comment... You seem to be going out of your way to work arround the abstraction solr provides, which isn't really a good idea in Solr, let alone in tests of solr. I smell a probable XY problem... http://people.apache.org/~hossman/#xyproblem ...can you explain what your *goal* is (indepentent of the approach you are taking to do it), and then maybe folks can make better suggestions about a better approach? -Hoss