Keith Seitz wrote:

Hi,

The attached patch should fix the mark problem with ByteBuffer.compact, which simply was not following the API specification.

I'm also attaching a simple jtreg testcase for this.

Keith Keith - I think Iris is going to get back to you shortly on this. One thing that needs to be figured out here is the implication for any existing applications that might depend on the current (broken) behavior. Unfortunately we have to worry about such things. In this case, if there is existing code that (perhaps unknowingly) depends on the current behavior then we might break it by a reset throwing an invalid mark exception in cases where it isn't thrown today. In the past we've had to introduce system properties to allow for "bug compatibility" with previous releases. I'm not saying we need it here but we need to think about it.

-Alan.