[HADOOP-15962] The buffer size is small when unpacking tar archives

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 3.3.0
• Component/s: common, util
• Hadoop Flags: Reviewed
• Flags: Patch
• Assignee: David Mollitor
• Reporter: David Mollitor
• Created: 30/Nov/18 23:31
• Updated: 28/Dec/18 11:49
• Resolved: 28/Dec/18 06:41

Description

Note sure if this code is even being used, but it implements a copy routing utilizing a 2K buffer.  Modern JVM uses 8K, but 4K should be minimum.  Also, there are libraries for this stuff.

FileUtil.java

    int count;
    byte data[] = new byte[2048];
    try (BufferedOutputStream outputStream = new BufferedOutputStream(
        new FileOutputStream(outputFile));) {

      while ((count = tis.read(data)) != -1) {
        outputStream.write(data, 0, count);
      }

      outputStream.flush();
    }

I also fixed a couple of check-style warnings.