[HADOOP-3440] TaskRunner creates a symlink with name 'null' if a file is added to DistributedCache without fragment

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 0.18.0
• Hadoop Flags: Reviewed
• Assignee: Devaraj Das
• Reporter: Abhijit Bagri
• Created: 23/May/08 15:12
• Updated: 08/Jul/09 16:52
• Resolved: 06/Jun/08 19:35

Description

When adding a file to DistributedCache by either :

1. DistributedCache.addCacheArchive(URI, conf) with URI not defining fragment(e.g. not as new URI("path#fragment") )
2. DistributedCache.addCacheFile(URI, conf) with URI not defining fragment(e.g. not as new URI("path#fragment") )
3. DistributedCache.addArchiveToClassPath(Path, conf)

and one sets DistributedCache.createSymlink, TaskRunner creates a link with name 'null' in the working directory.

I think the behavior should be that if fragment is null then no sym link should be created, adding in localizeCache() of DistributedCache the following

     boolean doSymlink = getSymlink(conf);
+    if(cache.getFragment() == null) {
+      doSymlink = false;
+    }
FileSystem fs = getFileSystem(cache, conf);