[Pharo-compiler] New Compiler - failing tests

Marcus Denker marcus.denker at inria.fr
Tue Apr 6 13:29:30 CEST 2010


On Apr 5, 2010, at 7:31 PM, Jorge Ressia wrote:

> Sure, no problem
> 
> I already saw what was it.
> 
> Basically a change I did a long time ago was removed.
> 
> Before we had "return self" but in order to do so there was some code
> for evaluating this in several places. I removed that since I thought
> that this issue should be decided in the optimization phase. So we
> have "push self" and then "return top".
> This is something we discussed some time ago, our objective was to
> keep the phases separated.
> 

Yes! So he idea would be to handle returnSelf like popIntoInstavar:
	-> the user just has storeIVar and pop at the level of IRBuilder
	-> the backend combines the pop into the preceeding IR when
             gnerating bytecode.

same pushSelf / ReturnTop --> return self.

> I would suggest keeping my change if there is no other point against.
> 
Nope. I think this is the way to go.

> Also I would go through the compilation tests to provide a base of
> green tests for other people to use and work on.

Yes! We just looked at the low-level test decompile IR/compile IR.
We will have a look at the others, soon... and keep them green ;-)

	Marcus


--
Marcus Denker  -- http://www.marcusdenker.de
INRIA Lille -- Nord Europe. Team RMoD.




More information about the Pharo-compiler mailing list