Comment 1 Jesper Moller 2013-03-28 13:45:06 EDT

Created attachment 229162 [details]
Work in progress

Working, but a bit rough:
 - ConstantPool sections for MethodHandles, MethodType, InvokeDynamic (but lacks caching)
  - Attribute section for BootStrapMethods
  - Adding of InnerClass reference for MethodHandles.Lookup
  - Implementation generateCode for ReferenceExpression
And it actually runs, as shown in the supplied test.

Yes, I know it lacks comment/disclaimer work.
Also lacks performance/polish in putting together InvokeDynamic constants from code generation.

Comment 2 Jesper Moller 2013-03-31 03:24:41 EDT

Created attachment 229192 [details]
Improved work in progress

Should cover all cases of reference expressions, static/non-static, with/without receiver.

Comment 3 Jesper Moller 2013-04-02 04:53:05 EDT

Created attachment 229224 [details]
Patch with functioning constructor references and generics

I'll leave this for now, as plumbing for lambda is needed anyway, and the design needs review.

Comment 4 Srikanth Sankaran 2013-04-06 06:07:53 EDT

After being mystified and wasting quite a few hours tearing my hair out,
I learnt that javap does not show the lambda method bodies unless invoked
with -private. It truly looked like magic that things would work at all
without the lambda method bodies being there in the class file :)

Comment 5 Jesper Moller 2013-04-06 07:08:49 EDT

(In reply to comment #4)
> After being mystified and wasting quite a few hours tearing my hair out,
> I learnt that javap does not show the lambda method bodies unless invoked
> with -private. It truly looked like magic that things would work at all
> without the lambda method bodies being there in the class file :)

Magic it is - I had to read up on invokedynamic a lot, too. You also need -v to get the BootStrapMethod attribute from javap.

Comment 6 Srikanth Sankaran 2013-04-17 09:37:49 EDT

Please do not post patches here anymore. Let us use this as the umbrella
bug for tracking code generation project and to hold high level discussions.
Patches should be attached to individual leaf level bugs that are blockers
for the current bug.

Comment 7 Srikanth Sankaran 2013-04-29 17:57:46 EDT

This is mostly cooked except for varargs and serialization support.
Jesper, if I prepare a cumulative patch, are you willing to review
and do some white box testing ? TIA,

Comment 8 Srikanth Sankaran 2014-01-06 09:43:41 EST

This umbrella bug has served its purpose. Resolving even though a couple
of work items are still open - they are tracked on their own ticket.