Current heap dump parser relies on class records to appear after the superclass records appear. This does not actually hold for hprof: we can emit the subclass records first, and only then emit the superclass records.
Therefore, current heap dump reader is broken when it encounters hierarchical classes. It would then underestimate the footprint of subclasses, because it does not record superclasses.
Therefore, current heap dump reader is broken when it encounters hierarchical classes. It would then underestimate the footprint of subclasses, because it does not record superclasses.
- links to
-
Review openjdk/jol/53