On 11/08/2011 19:31, Michael McMahon wrote:
On 11/08/11 17:27, Alan Bateman wrote:
Chris Hegarty wrote:
Here is a first stab at fixing this issue.
http://cr.openjdk.java.net/~chegar/7014860/webrev.00/webrev/
-Chris.
The spec clarification seems reasonable to me. Small typo "the
streams available method" to "the stream's available method".
Good to see the NIO socket adapter included in the test. One comment
on the test is that the Thread.dumpStack is a bit odd. It might be
better to pass in a test name into the test method and have that
printed by the failure message.
-Alan.
Just wondering what the motivation for changing this is? I can see the
logic in the change
all right, but has anyone actually encountered this problem (of an
inconsistency
between InputStream.available() and the number of bytes that can be
returned)?
This fix gives us consistent behavior across all platforms, and
clarifies the spec so that users know what to expect. Currently the
behavior on Linux and Windows is to return the amount of unattainable
data in the socket buffer. Solaris returns 0.
We have two bugs in the bug database (that I can find) for this:
6726928: SocketInputStream.available() method does not return 0 when
it reaches end of file on Linux
7014860: Socket.getInputStream().available() not clear for case that
connection is shutdown for reading
CR 7014860 is the bug being used to fix this issue.
-Chris.
- Michael
