On Tue, Sep 7, 2010 at 1:07 PM, Nicolas Cellier <
nicolas.cellier.aka.nice at gmail.com> wrote:

> About http://bugs.squeak.org/view.php?id=3448, the visible
> consequences are fixed by introduction of closure, but...
>
> Can someone explain why should ParagraphEditor & al catch
> OutOfScopeNotification at all ?
>

Looking at the Mantis case I'd say it was obsolete and ideally we would nuke
the entire class and its uses, but we need to change the part of
Encoder>>#encodeVariable:sourceRange:ifUnknown: that says

  (varNode isTemp and: [varNode scope < 0]) ifTrue:
  [^OutOfScopeNotification signal
 ifTrue: [action value]
ifFalse: [self notify: 'out of scope']].

to "do the right thing".  This isn't the right thing:

  (varNode isTemp and: [varNode scope < 0]) ifTrue:
  [^action value].

Because for an out-of-scope temp this will cause a raise of an
UndeclaredVariableWarning, which is wrong, which if uncaught will result in
a notification of "Undeclared", which is misleading.

Instead one fix might be to change Encoder>>encodeVariable: name to
something like
Encoder>>encodeVariable: name
^self encodeVariable: name
 sourceRange: nil
 ifOutOfScope:
 [:nodeOrNil|
 nodeOrNil
  ifNil: [self undeclared: name]
  ifNotNil: [self assert: (nodeOrNil isTemp and: [nodeOrNil scope < 0]).
      requestor notify: (nodeOrNil isArg ifTrue: ['block arg'] ifFalse:
['temp']),
    ' is out of scope']]

rename Encoder>>#encodeVariable:sourceRange:ifUnknown:
to Encoder>>#encodeVariable:sourceRange:ifOutOfScope: and have it pass in
either nil or the out-of-scope temp.

What would you do?

best,
Eliot


>
> Nicolas
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20100907/5153bb18/attachment.htm