* Unnamed modules should not have a lexical parent.

… class has the right parent.

* And try to partially fix AbstractGeneralSuperCallNode.

…om the classloader

* Copy and recompile to be sure to have no dependency for now.

update testing commands

Fix for issue 2050 on master: BigDecimal and nil multiplication

* We can use methods of Kernel and bind them on BasicObject:

> f = Kernel.instance_method :freeze
=> #<UnboundMethod: Kernel#freeze>
> b=BasicObject.new
=> #<BasicObject:0x62e2d8>
> is=Kernel.instance_method :frozen?
=> #<UnboundMethod: Kernel#frozen?>
> is.bind(b).call
=> false
> f.bind(b).call
=> #<BasicObject:0x47ef28>
> is.bind(b).call
=> true

…ton class.

* Fix bad formatting in frozenError().

clear out extra code and comments

* Fixes output of bin/ast -s.

* Simplify too much indices computations by using negative indices.
* Introduce an ArraySliceNode.

…etter splat support.

* As noted in the github issue comments, access protection
  checks seem to currently fetch the caller's self via the
  frame stack. This means that any scope that has instructions
  that requires access checks will force a frame push for that
  stack.

* For this patch, I have added a fix for attr assign instruction
  which seems to be the only instr that seems to directly call
  Helpers.invoke(..) which then calls the access protection methods
  in RubyClass.

  We need to audit if other call instructions require the same
  fixes.

* A better fix would be to fix Helpers.invoke(..) to accept an
  additional argument (caller self) which may be available at the
  call site without accessing the frame stack.

* Added a regression spec that fails without the fix in this patch.

Prepend current

… us!