RFR(M) 8208999 Some use of Klass* should be replaced by InstanceKlass*

Ioi Lam ioi.lam at oracle.com
Mon Aug 6 21:26:30 UTC 2018

Hi, please review this clean up


I changed the following APIs so that we can avoid many unnecessary type 
from Klass* -> InstanceKlass*.

     Klass::java_super()                    -> now returns InstanceKlass*
     InstanceKlass::local_interfaces()      -> Array<InstanceKlass*>*
     InstanceKlass::transitive_interfaces() -> Array<InstanceKlass*>*

I also modified APIs in SystemSictionary and SystemSictionary to return 
where possible.

Most changes are mundane, but here are some notable changes:

1. To save space, Klass::_secondary_supers (an Array<Klass*>*) may be 
    to the same array as InstanceKlass::_transitive_interfaces (an

    To make this happen, I added Array::as_const_array_of().
    See InstanceKlass::compute_secondary_supers() and array.h.

    I've added enough 'const' and asserts to catch mis-uses of this type
    of casting. If someone has a better suggestion, please let me know.

2. I changed Klass::_secondary_supers to 'const Array<Klass*>*' because
    of #1.

    There are many pointers returned by Klass/InstanceKlass that should be
    'const', as most callers should read them only without modifying, 
but that's
    outside the scope of this RFE.

3. SystemDictionary::resolve_super_or_fail() currently accepts the format
    of "Ljava/lang/Object;" for the "super_name" parameter. I am keeping the
    same behavior in my patch. However, I wonder if this actually correct
    per JVMLS. I filed JDK-8209030.

    See the call to resolve_instance_class_or_null_helper from 

Testing: hs tiers 1/2/3/4

- Ioi

More information about the hotspot-runtime-dev mailing list