Hmm. That would seem to be more consistent. I'll take another look and
likely file a follow-up bug since I had already pushed JDK-8231546 when
I saw this message.
Thanks for the careful review,
-Joe
On 9/26/2019 12:58 PM, Chris Hegarty wrote:
+1. Should the @serial javadoc tag be removed?
-Chris.
On 26 Sep 2019, at 19:27, Roger Riggs <[email protected]> wrote:
Looks fine, Joe
On 9/26/19 1:51 PM, Joe Darcy wrote:
Hello,
Another part of the cleanup of library serialization usage ahead of augmented
javac -Xlint warnings (JDK-8160675), the non-serializable instance fields of
serializable types in the java.prefs module should have their serial warnings
suppressed. The analogous issue in core libs is JDK-8231202.
For the type in question, while NodeChangeEvent is typed as serializable, both its
readObject and writeObject methods are written to unconditionally throw exceptions, so it
is not serializable in practice. The class does define a serialVersionUID, so the one
suspect field can be marked as "transient" without causing an effective change
in serialization support.
Patch below.
Thanks,
-Joe
diff -r e23e560afbcb
src/java.prefs/share/classes/java/util/prefs/NodeChangeEvent.java
--- a/src/java.prefs/share/classes/java/util/prefs/NodeChangeEvent.java Wed Sep
25 21:26:38 2019 -0700
+++ b/src/java.prefs/share/classes/java/util/prefs/NodeChangeEvent.java Thu Sep
26 10:48:34 2019 -0700
@@ -49,7 +49,7 @@
*
* @serial
*/
- private Preferences child;
+ private transient Preferences child;
/**
* Constructs a new {@code NodeChangeEvent} instance.
