[squeak-dev] Re: to be deployed Epaati version is out!

[karl]

Yoshiki Ohshima wrote:
>>   I'm playing with Epaati-10 a bit.  Entering
>> Grade2/Math/Unit4/IIM4_2_money identification.011.pr and coming back
>> (the instance of Project did get collected, but the accompanying
>> PasteUpMorph serving as its world along with all objects and players
>> are lingering.  Now, I'm (again) looking at the issue so hopefully I
>> get to something...
>>     
>
>   Just a progress report, but the issue is basically around
> #rootsIncludingPlayers not finding all classes, and the problem is
> caused by a project that has scripts that reference to an object that
> was trashed.  Namely,
>
>   - You created object A and object B.
>   - You wrote a script C at object B that refers to object A
>     (This creates the uniclass for B).
>   - You wrote a script D at object A that refers to object B.
>     (This creates the uniclass for A).
>   - You dismissed/trashed object B.
>   - The project was saved.
>
> What happens is that to keep the script D running and project working,
> the system exports the object B into the saved project as well.  But
> because it is trashed, it is not "in the world", but referenced from
> the scripts.
>
>   Epaati loads such a project, and upon exiting the project, it tries
> to remove the project.  From #okToChangeSilently,
> #rootsIncludingPlayers is called to find the uniclasses used in the
> project.  But the logic only looks at the objects in the world, and
> overlook the object B and the B's uniclass.
>
>   Because B has a script that refers to A, pretty much everything in
> the project is kept because the world is reachable through A's owner
> chain.
>   

Yes, there should be a similar mechanism as when you delete referenced 
method in the Browser, that open or list the scripts that references the 
deleted script or player.


>   I still think Etoys/Smalltalk is almost suitable for what you are
> doing, but loading and unloading a lot of project in a session wasn't
> a typical use case.  In a sense Epaati is stretching it.  But it is
> fixable fortunately.
>
>   One thing we definitely should do is to make #rootsIncludingPlayers
> better.  I can think of a few different ways.  One thing you should do
> is revisit your projects and make sure that every object refered to
> from the project to "live" in the project.
> IIM4_2_money identification.011.pr, for example, has quite a few of such objects.
>
>   To check these guys, open a workspace in a fresh epaati.image, and
> evaluate:
>
>     old := PasteUpMorph allInstances.
>
> Then load IIM4_2_money identification.011.pr and come back.  In the
> same workspace evaluate:
>
>     new := PasteUpMorph allInstances.
>     new := (new copyFrom: old size + 1 to: new size).
>     new := new select: [:e | e knownName = 'page'].
>
> and look at the submorphs of these pages bound to new.
>
> -- Yoshiki
>
>
>   
We get the environment tested thoroughly here and that is really good.

Karl