Yep, I agree that SUIDs must be the same only for classes with the documented serialized form.
SY, Alexey 2007/6/15, Zakharov, Vasily M <[EMAIL PROTECTED]>:
Dmitry, As far as I know, Swing is not supposed to be serialization-compatible not only between implementations, but even between different versions of the same implementation. You may see that, though Swing classes are Serializable, they do not have "Serialized form" link in the API specification, and are not listed at the "Serialized form" page. In this situation having different serialVersionUID in Harmony than in RI looks like a feature, not a bug - it makes sure that no deserialization would be attempted in situation where compatibility is not supposed. Also, Swing classes do not have serialVersionUID value specified in the API specification, so we do not violate the spec by having a different value than RI in our Swing implementation. Note: this justification is Swing-specific and is only applicable to Swing. Vasily Zakharov Intel ESSD -----Original Message----- From: Dmitry Irlyanov [mailto:[EMAIL PROTECTED] Sent: Friday, June 15, 2007 5:01 PM To: [email protected] Subject: [classlib] serialVersionUID justification Hi, Harmony differs from RI implementation not only in methods implementation but also in class fields. The problem is correctly stated in the JIRA issue ( https://issues.apache.org/jira/browse/HARMONY-4187 ) that says that values of serialVersionUID differ for Harmony and RI. You can see it at http://www.kaffe.org/~stuart/japi/htmlout/h-jdk15-harmony.html , but I'm totally confused what to do if serialVersionUID equals smth in jdk 1.5 and smth else in Harmony. Should I prepare the patch that changes the values to match/be equal? _______________ With Best Regards, Irlyanov Dmitry
