6516099: InputStream.skipFully(int k) to skip exactly k bytes

Brian Burkhalter brian.burkhalter at oracle.com
Fri Nov 30 19:55:08 UTC 2018

Loathe though I am to resurrect this thread, one problem arose after testing with assertions enabled which I had neglected previously to do. A few small changes to test/jdk/java/io/InputStream/Skip.java are required. A delta versus webrev.08 (link at bottom) is at


The changes are as follows (line numbers in the new version):

1. L154

         in.setState(-1, 100);
-        dotestExact(in, pos, streamLength, n, true, false);
+        dotestExact(in, pos, streamLength, -1, true, false);

Pass “-1” as the number to skip instead of “n.”

2. L159

         in.setState(n + 1, 100);
         dotestExact(in, pos, streamLength, n, true, false);
+        pos += n + 1;

Update the test-tracked position according to the number actually skipped, which is “n + 1” although only “n” bytes were requested. This sub-test causes the internal call to “skip()” to be “skip(n+1)” which will return “n+1” which intentionally provokes an IOException and leaves the stream in an inconsistent state as documented. The call to “skip(n+1)” does however really skip “n+1” bytes so the test-tracked position must be updated accordingly.

3. L214

-    public long position() { return readctr == endoffile ? EOF : readctr; }
+    public long position() { return readctr; }

The position should be returning the actual offset which for EOF is the length of the stream, not -1.

With the foregoing changes the test passes.



> On Nov 28, 2018, at 9:51 AM, Daniel Fuchs <daniel.fuchs at oracle.com> wrote:
> Looks good to me Brian.
> I never knew whether positive meant >= 0 or > 0 anyway ;-)
> best regards,
> -- daniel
> On 28/11/2018 17:38, Brian Burkhalter wrote:
>> http://cr.openjdk.java.net/~bpb/6516099/webrev.08/

More information about the core-libs-dev mailing list