(original) (raw)


Hi,

I changed the implementation according to Brad's comments. I am reposting the output of hg diff src/share/classes/java/net/DatagramSocket.java since I don't have an openjdk account:

bash-3.00$ hg diff src/share/classes/java/net/DatagramSocket.java
diff --git a/src/share/classes/java/net/DatagramSocket.java b/src/share/classes/java/net/DatagramSocket.java
\--- a/src/share/classes/java/net/DatagramSocket.java
+++ b/src/share/classes/java/net/DatagramSocket.java
@@ -176,13 +176,7 @@ class DatagramSocket implements java.io.
public DatagramSocket() throws SocketException {
// create a datagram socket.
createImpl();
\- try {
\- bind(new InetSocketAddress(0));
\- } catch (SocketException se) {
\- throw se;
\- } catch(IOException e) {
\- throw new SocketException(e.getMessage());
\- }
\+ bind(new InetSocketAddress(0));
}


Thanks,
Kurchi



On 7/22/2011 7:25 AM, Michael McMahon wrote:
On 22/07/11 14:55, Alan Bateman wrote:
Michael McMahon wrote:
But, bind() already closes the impl internally before throwing the exception.
I was wondering about that and whether this is a bug. Suppose someone creates an unbound DatagramSocket and then attempts to bind it to a port. If the bind fails (say port already in use) then it may be surprising that they can't retry with a different port. Should this be specified? I see there are cases such as the security exception where it doesn't the close the impl.

\-Alan.
It doesn't seem to be specified either way, though it does seem to be inconsistent.
I'd be wary about changing the behaviour though
unless there was a strong justification (more than a compiler warning :) )

\- Michael.