> On 8 Mar 2016, at 00:24, Uwe Schindler <uschind...@apache.org> wrote: > > Hi Steve, > > Thanks for the quick fix! I am not able to test this on the short term, but > I trust you that Lucene builds now.
I built it successfully a few times from scratch (downloading half the internet :-) ). > I am a bit nervous, because it does not explain the Ivy issues, but I will > try to create some test cases with relative jar:-URL resolving tomorrow. Thanks. Note that the resource URLs produced from the class loader should no longer have the #runtime fragment, unless those resources are from an MR-JAR. #runtime is the signal to the URL protocol implementations to process as runtime versioned resource. > This may help with resolving the problems in build 110. > > I just want to make sure, that the following also works: > - Get URL from classloader to a resource file > - resolve a relative file against this URL and load it by URL > (this is common pattern for parsing XML resources from JAR files that refer > relatively to other resources in same JAR file by href) > If you have a small test project you can share we can give it a test run in the interim. It could be that the URL resolving mechanism worked incorrectly with a #fragment in the way (especially that mechanism operated directly on the characters of the URL). > Keep me informed when build 109 is downloadable. > Will do. Paul.
