Bill Pugh writes:

There is a bug in java.io.InputStream.skip().

Stuff related to this has been reported before, but I think we can make 
a strong
case that the behavior I'm reporting is a bug that needs to be fixed.

The spec for InputStream.available() says:


>> Returns the number of bytes that can be read (or skipped over) from  
>> this input stream without blocking by the next caller of a method for  
>> this input stream.


However, BufferedInputStream.skip(long) can decide to skip less than 
the number
of bytes that were reported as being available to skip.

Here is a test case:

public class SkipBug {

		public static void main(String args[]) throws IOException {
			int bufferSize = 20;
			int skipRequestSize = bufferSize;
			ByteArrayInputStream bytes = new ByteArrayInputStream(new byte 
[bufferSize*10]);
			BufferedInputStream buffered = new BufferedInputStream(bytes, 
bufferSize);
			buffered.read();
			int avail = buffered.available();
			long skipped = buffered.skip(skipRequestSize);
			assert skipped >= Math.min(skipRequestSize, avail)
				: "Skipped only " + skipped + " bytes but "
							+  avail + " were available";
			
		}
}

The call to available will report that 199 bytes are available to skip, 
but the call to skip 20 bytes will only skip 19.

There is a fair bit of code that depends on being able to read or skip 
as many
bytes as are returned by available(), and that doesn't work for 
BufferedInputStream and skip().

I think the spec for InputStream should be clarified to say that it is 
intended/desirable/required that read(byte[], int, int) and skip(long) 
will not read/skip over fewer bytes
than are available unless fewer bytes than that are requested. The spec 
for available() clearly seems to imply this.

   In others words,
		min(bytes available, bytes requested) <= bytes read/skipped <= bytes 
requested

Note that if an InputStream.available() returns 0, this has no impact.

This is related to bug 4254082
	http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4254082
but that bug doesn't mention available().

We've actually got a bug detector in FindBugs that flags all uses of 
skip where the return value is not checked on a BufferedInputStream due 
to this flaw in the Java i/o libraries.
I'd like to be able to change that so it doesn't warn about cases in 
which the programmer
has checked available(). 


###@###.### 2004-12-04 22:18:38 GMT