Andrew John Hughes <A.Hughes <at> dcs.shef.ac.uk> writes: > There seem to be an awful lot of 'missing' classes on this reverse > JAPI run. Presumably, these are not missing, but we are wrongly > exposing them in GNU Classpath?
That's pretty much right. You have to read the results backwards in this report. An item that's "missing in jdk15" is a class that's public or protected in Classpath that either doesn't exist at all in JDK1.5 or is default-access or private. These items should be made default-access or private in Classpath too, or moved to a gnu.* package, or removed entirely if they're unused. Simplest way to eliminate the Japi errors would be to just make them default-access and make sure everything still compiles, but (just like with stubs) it might be better not to do this without figuring what the "right way" is in each case. Another note for anyone interested in tackling these: Japi considers all methods of a final class final, so in several cases in the current report a class is reported as incorrectly nonfinal and then the same error appears for all of its methods. Fixing the class to be final is all that's needed to eliminate all those errors. Stuart.
