[Vm-dev] Proxies and primitives

[Igor Stasenko]

On 17 January 2012 12:35, Stefan Marr <smalltalk at stefan-marr.de> wrote:
>
>
> On 17 Jan 2012, at 11:52, Mariano Martinez Peck wrote:
>
>> Hi guys. I am experimenting with something that creates proxies automatically and become existing objects to those proxies. The objects I become with proxies are selected kind of automatically and they can be instances of any class. I have a list of classes that I don't become their instances since it seems I can easily crash the VM (such as SmallIntegers, Characters, Symbols, etc).  Now...with mi proxy implementation I can intercept almost all messages, so I don't have a problem with that. What I am thinking now is the problem of proxies that are passed as arguments to primitives. My question is if there can be primitives that assumes the type of the argument and do something directly. So if I send a proxy that doesn't respect the expected shape...it can answer something wrong or even crash. Another possibility is that the primitive fails normally (primitiveFailure) but then that there is no smalltalk code after the primitive invocation (I guess not all methods that use a primitive has the correct smalltalk code after).
>> So.... Is this possible?  do you have an example?
>
> Example: the perform with super primitive. (if the expected array aint an array, you can expect at least your stack to get messed up on Cog)
>
> My understanding is currently that primitives do not guarantee you anything if you violate their contract.
>
> You can easily crash Cog since Eliot makes that assumption.
>
> So, what you are doing is inherently dangerous, and requires you to adapt all primitives to handle these cases.
>
> As I said, that is my current understanding. I think you should refer to the blue book and see what is discussed there with respect to primitives and guarantees.
>


yes, that's the most compelling part of VM: you can relatively easy
change semantics of 'getting reference to an object' and change
contracts,
but then when you meet primitives, you realizing that any of such
change could break a contracts with primitives.
And if you start fixing them, it is really hard to tell if you changed
them all and didn't missed one or two.. because there's a lot of them
:(


> Best regards
> Stefan
> --
> Stefan Marr
> Software Languages Lab
> Vrije Universiteit Brussel
> Pleinlaan 2 / B-1050 Brussels / Belgium
> http://soft.vub.ac.be/~smarr
> Phone: +32 2 629 2974
> Fax:   +32 2 629 3525
>



-- 
Best regards,
Igor Stasenko.