Re: agents / TUPLE inheritance

Hi Raphael,

I think you may have encountered a consequence of a difference of 
opinion described in this paper by Philippe Ribet:

http://www.jot.fm/issues/issue_2004_04/article7

I use the ISE compiler, and on several occasions I have encountered the 
problem that the article seeks to fix. For this reason, I have tended 
to agree with the SmartEiffel approach to agent conformance.

Your problem is interesting, though. It seems to demonstrate that the 
ISE/ECMA approach also has advantages.

- Peter Gummer


On 08/04/2005, at 3:58 AM, Raphael Mack wrote:

> Hi,
>
> I tired to compile the following code:
>
> class A
>
> creation
>     make
>
> feature
>     make is
>         local
>             t: T1
>         do
>             create {T1}t
>             a := agent g(?)
>             a.call([t])
>         end
>
>     g(i: T1) is
>         do
>             io.put_string(i.out)
>         end
>
>     a: ROUTINE[ANY, TUPLE[TT]]
>
> end
>
>
> with T1 and TT:
> class TT
> end
>
>
> class T1
> inherit TT
>         redefine
>                 out
>         end
> creation
>         default_create
> feature
>         out: STRING is
>                 do
>                         Result := "T1..."
>                 end
> end
>
> And I get
> ****** Fatal Error:  Bad assignment (VJAR). Cannot assign `agent
> g(?)' which is of type PROCEDURE[ANY,TUPLE[T1]] into `a' which
> is of type ROUTINE[ANY,TUPLE[TT]].
> Line 12 columns 4, 9 in A (/home/r/eiffel/se_test/a.e):
>          a := agent g(?)
>          ^    ^
> ------
>
> For me, this seems to be perfectly fine Eiffel (and it works with ise) 
> - what's wrong with it?
>
> Thanks,
> Raphael
>