Michael Koch writes: > On Sat, Mar 25, 2006 at 10:06:15AM +0000, Andrew Haley wrote: > > Tom Tromey writes: > > > >>>>> "Stuart" == Stuart Ballard <[EMAIL PROTECTED]> writes: > > > > > > Stuart> Perhaps in the same spirit as the "throws > > NotImplementedException" > > > Stuart> hack, we could add a "NotReallySerializable" interface that > > classes > > > Stuart> whose svuids should be ignored could implement... > > > > > > I don't think we could do that, as a new 'implements' is user-visible. > > > An annotation would be ok, I suppose. > > > > Could we not, instead, automatically fix the SerialVersionUID > > mismatches? After all, if we can automatically determine that they're > > different, we can automatically fix them, can't we? > > The question is what the fix is? In every service release of SUNs JDK > they change some of the serialVersionUIDs. Whenever you run japi locally > against another service release of SUN you will get other differences.
OK, so the classes that are not stable with respect to SerialVersionUID don't need to be fixed. But my point stands: those that should have the SerialVersionUID mismatches fixed can surely be fixed automatically, either by correcting or adding a SerialVersionUID field. DO we even know for sure which SerialVersionUIDs should be declared? Andrew.