On Tue, 1 Sep 2026 10:25:09 GMT, Oli Gillespie <[email protected]> wrote:
> Do you mean to represent the filler array as FillerElement[]...
Sth along that line, so that jmap and heap-dump provide consistent views. Since
this filler-obj/array is special, it doesn't need to behave like a real array,
with proper array-len and per-element size.
> it's just that the heap dumper sees...
Impl wise, how about another branch in `HeapObjectDumper::do_object`?
} else if (o->is_filler()) {
// dump filler with its friendly klass-name.
} else if (o->is_typeArray()) {
...
}
> ... I think exposing fillers in heap dump brings more problems than it solves
If they were never exposed, it's probably true. Now that there are multiple
observers (jmap, jstat, heap-dump, etc) into the heap, I am leaned towards
presenting heap as-is, showing the presence of filler-obj/array and annotate
them differently so that fillers can be distinguished from ordinary int-array.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/32542#issuecomment-5494198991
