On Mar 28, 2011, at 11:28 AM, Mariano Martinez Peck wrote:

Hi folks. I would like to know which are the messages that are not really executed, thus not being intercepted by proxies.
So far, I know:

1)  
those are inlined by the compiler...

yes.

2)  Special shorcut bytecodes that are in "Smalltalk specialSelectors" / "ParseNode classVarNamed:  'StdSelectors'"  / "Smalltalk specialObjectsArray at: 24"

Here I have a question...if I understand correctly, in this list of selectors there are 3 cases:

2.1)  methods may or may not be executed depending on the receiver/arguments. For example
1 + 2  will never send the message #+ to 1.
but 1 + 'mariano'   will do.


yes.

2.2) methods that are NEVER sent,   like #class and #==

yes

2.3) methods that are ALWAYS sent, like #new, #next, #nextPut:, #size, etc.

So...did I understand correclty this part?

I think so, yes.


are there more cases where a message may not be really executed?

You could see the block creation bytecode as a problematic case.
It's not a message, so you can't trap it. 

But then when we are on that level, the same is with iVar access. Or temp access.


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