Resolves GSA-681 / Compressed VCF.gz output is too big because of unnecessary call to flush().

-- Now compressed output VCFs are properly blocked compressed (i.e., they are actually smaller than the uncompressed VCF)

protected/java/test/org/broadinstitute/sting/gatk/walkers/genotyper/UnifiedGenotyperIntegrationTest.java

@@ -96,7 +96,7 @@ public class UnifiedGenotyperIntegrationTest extends WalkerTest {
     //
     // --------------------------------------------------------------------------------------------------------------
 
-    private final static String COMPRESSED_OUTPUT_MD5 = "5b8f477c287770b5769b05591e35bc2d";
+    private final static String COMPRESSED_OUTPUT_MD5 = "3eba6c309514d1e9ee06a20a112b68e6";
 
     @Test
     public void testCompressedOutput() {

public/java/src/org/broadinstitute/sting/utils/variantcontext/writer/VCFWriter.java

@@ -248,7 +248,8 @@ class VCFWriter extends IndexingVariantContextWriter {
             }
             
             mWriter.write("\n");
-            mWriter.flush();  // necessary so that writing to an output stream will work
+            // note that we cannot call flush here if we want block gzipping to work properly
+            // calling flush results in all gzipped blocks for each variant
         } catch (IOException e) {
             throw new RuntimeException("Unable to write the VCF object to " + getStreamName(), e);
         }