[hibernate-issues] [Hibernate-JIRA] Commented: (BVTCK-11) Wrong usage of context classloader to discover resources in META-INF/services

[ http://opensource.atlassian.com/projects/hibernate/browse/BVTCK-11?page=c... ] David Jencks commented on BVTCK-11: ----------------------------------- Sorry we (geronimo) let this drop for a while since we found a workaround that is now causing more serious problems. IIUC everyone agrees that the current test is invalid. Can it be excluded right now? I think the proposed solution would work with geronimo. However, I don't think it tests the assertion and I'm pretty sure the proposed default spi implementation is not what you want (what loader is actually returned?). The spec says the services file has to be in a jar file. If you mean this, then the test should be supplied the location of said jar file, verify its a jar file and not a directory or some other kind of classloader backing (e.g. a database), look inside just that jar file, and get the services file out. If its really a jar file rather than say an osgi bundle with bundle-classpath or osgi fragment, then there can only be one services file in the jar. If you don't actually mean the services file needs to be in a jar file, just accessible through a classloader, then the proposed pull request is probably ok although the spec needs adjustment. After you clarify this we can try out the proposal. Meanwhile we'd like the test excluded. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira