Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
1.4.1, 1.9
Description
BZip2CompressorInputStream.read(buffer, offset, length) returns -1 when given an offset equal to the length of the buffer.
This indicates, not that the buffer was full, but that the stream was finished.
It seems like a pretty stupid thing to do - but I'm getting this when trying to use Kryo serialization (which is probably a bug on their part, too), so it does occur and has negative affects.
Here's a JUnit test that shows the problem specifically:
@Test
public void testApacheCommonsBZipUncompression () throws Exception {
// Create a big random piece of data
byte[] rawData = new byte[1048576];
for (int i=0; i<rawData.length; ++i) {
rawData[i] = (byte) Math.floor(Math.random()*256);
}
// Compress it
ByteArrayOutputStream baos = new ByteArrayOutputStream();
BZip2CompressorOutputStream bzipOut = new BZip2CompressorOutputStream(baos);
bzipOut.write(rawData);
bzipOut.flush();
bzipOut.close();
baos.flush();
baos.close();
// Try to read it back in
ByteArrayInputStream bais = new ByteArrayInputStream(baos.toByteArray());
BZip2CompressorInputStream bzipIn = new BZip2CompressorInputStream(bais);
byte[] buffer = new byte[1024];
// Works fine
Assert.assertEquals(1024, bzipIn.read(buffer, 0, 1024));
// Fails, returns -1 (indicating the stream is complete rather than that the buffer
// was full)
Assert.assertEquals(0, bzipIn.read(buffer, 1024, 0));
// But if you change the above expected value to -1, the following line still works
Assert.assertEquals(1024, bzipIn.read(buffer, 0, 1024));
bzipIn.close();
}