jay vyas

@childsb ~~~~~~~~~~ see below ~~~~~~~~~~~~ $ git status# On branch branch-2.2 nothing to commit, working directory clean ``` $ cat `find ./ -name FileSystem.java` | grep -A 2 "String getScheme()"...

To clarify brad's comment: IT is true that the bug can be thought of as existing in the RR implementation in mapred version of MultiFileWordCount, and not necessarily a bug...

The offending code in RR for MultiFileWordCount violates the access pattern, here is why. (i agree its not the right way to use a stream). But in any case, here...

From ./src/examples/org/apache/hadoop/examples/MultiFileWordCount.java in hadoop 1.2: 151 public long getPos() throws IOException { 152 long currentOffset = currentStream == null ? 0 : currentStream.getPos(); 153 return offset + currentOffset; 154 }...

Alas, your right ! The problem is that currentStream is not null. However, the InputStreamReader IS set to null. That is the problem :) !

does it make sense now? The currentStream still exists, but the overlying wrapper (the Buffer) SETs the input stream reader to null. Thus, when the input stream reader is accesssed,...

https://issues.apache.org/jira/browse/MAPREDUCE-5572

simple. just do `hadoop fs -ls /` and you should see all the directories in gluster's root.

hi @mahsa-frj . 0) There are terasort, teragen, and so on. Those are available as part of your hadoop distribution in a jar file, and are also curated in apache...

Can you add a unit test for this ?