On 02/13/2014 02:31 AM, Alan Bateman wrote:
On 13/02/2014 08:00, Joe Darcy wrote:
Hello,
Please review the largely tedious but straightforward fix for
JDK-8034169: Fix serial lint warnings in javax.swing
http://cr.openjdk.java.net/~darcy/8034169.0/
With the change, all of java.swing and it subclasses compile cleanly
with the lint warning enabled.
Generally the serial warnings were address by an
@SuppressWarnings("serial") annotation; the annotation included a
comment about the justification for the suppressions.
In a small percentage of cases, a serialVersionUID field was added.
When such a field was added, the serialver computation was checked
for consistency on JDK6 and JDK 8.
(Note that the javax.swing.text.html.parser.Element type has actually
different serialVersionUIDs on JDK 6 and 8.)
I skimmed through the changes and I don't see anything obviously
wrong. However I am puzzled by some of the inconsistencies. For
example in javax.swing.event then some of the events have the warning
but some do not, they are examples in a few other packages too.
If a superclass had the standard swing serialization disclaimer, I put
@SuppressWarning("serial") on its subclasses. This covered most of the
cases. If one of the non-transient fields had teh swing disclaimer, I
added a note that that effect, and if the class was neither public nor
protected, I generally added a "JDK-implementation class" disclaimer.
This caught my eye because of the addition of SUID to public types
when I assumed they would have the warning and so have
@SuppressWarnings("serial"). Maybe this needs someone familiar with
the history in this area to say whether it is true that some of the
public types can be serialized in the compatible way across releases.
For exception types, types not meeting the criteria above (especially if
they haven't been modified in many years), I added a SUID, but those
cases were relatively few.
Hope this explanation helps; thanks for the review,
-Joe
