Hadoop: time conversion between Ceph and Hadoop;

ditch filling in the uid/gid since it's meaningless.

diff --git a/src/client/hadoop/ceph/CephFileSystem.java b/src/client/hadoop/ceph/CephFileSystem.java
index da899a4d12aa6d6195b577c49c7f620071d7677a..e38ae66533927ef61e6c797ad0dda7fd175f68fb 100644 (file)
--- a/src/client/hadoop/ceph/CephFileSystem.java
+++ b/src/client/hadoop/ceph/CephFileSystem.java
@@ -239,16 +239,22 @@ public class CephFileSystem extends FileSystem {
   public FileStatus getFileStatus(Path p) throws IOException {
     debug("getFileStatus:enter with path " + p);
     Path abs_p = makeAbsolute(p);
+    //sadly, Ceph doesn't really do uids/gids just yet, but
+    //everything else is filled
     // For the moment, hardwired replication
     int replication = 2;
     FileStatus status;
     Stat lstat = new Stat();
     if(ceph_stat(abs_p.toString(), lstat)) {
+      debug("getFileStatus: mod_time is " + lstat.mod_time);
       status = new FileStatus(lstat.size, lstat.is_dir, replication,
-                             lstat.block_size, lstat.mod_time, lstat.access_time,
+                             lstat.block_size,
+                             //these times in seconds get converted to millis
+                             lstat.mod_time*1000,
+                             lstat.access_time*1000,
                              new FsPermission((short)lstat.mode),
-                             new Integer(lstat.user_id).toString(), 
-                             new Integer(lstat.group_id).toString(),
+                             null,
+                             null,
                              new Path(fs_default_name+abs_p.toString()));
     }
     else { //fail out