On Dec 4, 2013, at 10:02 PM, Clément Bera <bera.clement@gmail.com> wrote:
It didn't halt on your machine because you didn't copy and paste the code that halts (i guess I wrote too much code in the mail so it was not clear). You copied the code that I explicitly said that doesn't halt because the error is raised at compilation time.
I saw it do not worry. I was just shocked by the state of the exceptions.
This code halts (copy, paste and run it you will see):
[ OpalCompiler new evaluate: 'AnUnknownClass subclass: #Adaptor1Example instanceVariableNames: ''customers accountID address name phoneNumber'' classVariableNames: '''' poolDictionaries: '''' category: ''Examples-Cookbook''' ] on: OCSemanticWarning do: [ :e | e halt ]
We can discuss this another day when you have time because seemingly my mail was not clear (you said you couldn't trap the semantic warning). Now I'm really sick today so we can discuss friday or next week.
rest well
2013/12/4 Stéphane Ducasse <stephane.ducasse@inria.fr>
Ok I will check something on UndefinedObject to start with. Then it would be fun to prototype a kind of strategy hiearchy on the side (extracted from also) behavior of OCSemanticsError It was nice to see it and since now I'm ok to have a look at the compiler code it was interesting.
The current design is motivated mostly by backward compatibility, not anything else.
I see :) I would really to see an headless, interactive and scriptable scenarios to be exercised. So this is good it will get better :) Stef