[ https://jira.jboss.org/browse/JBVFS-166?page=com.atlassian.jira.plugin.sy... ] Ales Justin reassigned JBVFS-166: --------------------------------- Assignee: Ales Justin (was: John Bailey) -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://issues.jboss.org/browse/JBVFS-166?page=com.atlassian.jira.plugin.... ] Ales Justin edited comment on JBVFS-166 at 5/7/12 9:30 AM: ----------------------------------------------------------- @Chris: how exactly do you reproduce this? This is my test atm: public void testAppAsLink() throws Exception { if (supportSymlinks() == false) return; URL dir = getResource("/vfs/symlink"); CombinedVFSCache cache = new CombinedVFSCache(); VFSCacheFactory.setInstance(cache); cache.setPermanentRoots(Collections.<URL, ExceptionHandler>singletonMap(dir, null)); VirtualFile root = VFS.getRoot(dir); VirtualFile app = root.getChild("app.jar"); VirtualFile clazz = app.getChild("org/jboss/test/vfs/support/CommonClass.class"); Assert.assertNotNull(clazz); URL url = clazz.toURL(); VirtualFile vf1 = VFS.getRoot(url); Assert.assertNotNull(vf1); VirtualFile vf2 = VFS.getRoot(url); Assert.assertNotNull(vf2); } Skywalker:symlink alesj$ ls -la total 8 drwxr-xr-x 3 alesj staff 102 May 7 15:16 . drwxr-xr-x 7 alesj staff 238 May 7 14:41 .. lrwxr-xr-x 1 alesj staff 17 May 7 15:16 app.jar -> ../test/outer.jar was (Author: alesj): @Chris: how exactly do you reproduce this? This is my test atm: public void testAppAsLink() throws Exception { if (supportSymlinks() == false) return; URL dir = getResource("/vfs/symlink"); CombinedVFSCache cache = new CombinedVFSCache(); VFSCacheFactory.setInstance(cache); cache.setPermanentRoots(Collections.<URL, ExceptionHandler>singletonMap(dir, null)); VirtualFile root = VFS.getRoot(dir); VirtualFile app = root.getChild("app.jar"); VirtualFile clazz = app.getChild("org/jboss/test/vfs/support/CommonClass.class"); Assert.assertNotNull(clazz); URL url = clazz.toURL(); VirtualFile vf1 = VFS.getRoot(url); Assert.assertNotNull(vf1); VirtualFile vf2 = VFS.getRoot(url); Assert.assertNotNull(vf2); } Skywalker:symlink alesj$ ls -la total 8 drwxr-xr-x 3 alesj staff 102 May 7 15:16 . drwxr-xr-x 7 alesj staff 238 May 7 14:41 .. lrwxr-xr-x 1 alesj staff 17 May 7 15:16 app.jar -> ../test/outer.jar -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://issues.jboss.org/browse/JBVFS-166?page=com.atlassian.jira.plugin.... ] Chris Sams commented on JBVFS-166: ---------------------------------- @Ales The permanent root should be a symlink and then something within it also should be a symlink whose resolved location does not include the permanent root's resolved path. For example: /vfs/symlink/im_a_symlink_too.jar /vfs/symlink can be the root and resolve to /tmp im_a_symlink_too.jar can resolve to /othertmp/the_real_jar.jar /tmp always goes in the permanent cache regardless of whether forceCanonical is true (from the property editor canonicalizing the path). If forceCanonical is true, subsequent accesses of /vfs/symlink/im_a_symlink_too.jar resolve to /othertmp/the_real_jar.jar, which isn't found in either the permanent or real cache. When adding the path to the cache after the miss, the uncanonicalized version is used, so a cache miss occurs every time after that, too. Now suppose forceCanonical is false. Then the path /vfs/symlink/im_a_symlink_too.jar is used when accessing a resource, and it gets added to the IterableTimedCache, which times out every 30 minutes. If the contents of the archive are exploded in memory, then you might not notice unless you update the jar and then wait for the timeout. There's an additional complexity when the archive is a deployed WAR file because the AbstractWarDeployer does the following on deployment: VFSDeploymentUnit vfsUnit = (VFSDeploymentUnit)unit; VirtualFile root = vfsUnit.getRoot(); VirtualFile unjared = VFSUtils.unjar(root); This is a problem because the TempInfo for the WAR is attached to the FileSystemContext of the deployment directory, which is always missed when accessing resources in the WAR. Instead those resources are loaded either directly from the WAR (every time if forceCanonical is true) or from memory (which gets refreshed every 30 minutes because of the IterableTimedCache timeout). Regardless, they're not coming from tmp as they should. I think that fixing the symlink resolution problems will fix the problem with the WAR also because all of the paths used should be consistent. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira

[ https://issues.jboss.org/browse/JBVFS-166?page=com.atlassian.jira.plugin.... ] Chris Sams commented on JBVFS-166: ---------------------------------- @Ales An updated version of the test and some code that causes it to succeed is in the JBVFS-166 branch of https://github.com/csams/jboss-vfs.git I've sent you a PR. As a sanity check, in addition to the testsuite, I've taken the jboss-vfs.jar from that branch and tested a previously failing scenario in my installation of EAP 5.1.2. It now works as expected. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.jboss.org/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira