[IO-428] BOMInputStream.skip returns wrong count if stream contains no BOM

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.0.1
• Fix Version/s: 2.5
• Component/s: Streams/Writers
• Assignee: Kristian Rosenvold
• Reporter: Stefan Gmeiner
• Created: 12/Mar/14 13:10
• Updated: 23/Jun/15 06:00
• Resolved: 23/Jun/15 06:00

Description

If the skip method of BOMInputStream is used on a stream without a BOM, skip returns the wrong number of bytes (n - max(BOM-length)). This can lead to problems if the return value is evaluated for example from guava ByteStreams.skipFully.

BomTest.java

public class BomTest {

	private static InputStream createInputStream(boolean addBOM) {
		ByteBuffer bb = ByteBuffer.allocate(64);
		if (addBOM) {
			// UTF-8 BOM
			bb.put(new byte[] { (byte) 0xEF, (byte) 0xBB, (byte) 0xBF });
		}
		bb.put((byte) 0x31);
		bb.put((byte) 0x32);
		bb.put((byte) 0x33);
		return new ByteArrayInputStream(bb.array());
	}
	
	public static void main(String[] args) throws IOException {
		BOMInputStream is1 = new BOMInputStream(createInputStream(true));
		assertEquals(2, is1.skip(2));
		assertEquals((byte) 0x33, is1.read());
		
		BOMInputStream is2 = new BOMInputStream(createInputStream(false));
		assertEquals(2, is2.skip(2)); // fails here - skip returns 0
		assertEquals((byte) 0x33, is2.read());
	}
	
}

I catched this bug in 2.0.1, but as far as I can see on the source 2.5 is still affected.

I suggest the following change to the skip method:

BOMInputStream.java

    public long skip(long n) throws IOException {
    	int skipped = 0;
        while ((n > skipped) && (readFirstBytes() >= 0)) {
            skipped++;
        }
        return in.skip(n - skipped) + skipped;
    }