Christoph Thiede wrote
> Hi Jaromir,
> 
>> > \2: This was indeed a slip because I forgot to update the image. I have
>> moved my patch to #findNextHandlerContext - it makes the method robust
>> against bottom-contexts that do not have a sender (i.e., sender is nil).
>> 
>> The changeset still seems to have the old version of
>> #runUntilErrorOrReturnFrom: and #nextHandlerContext nixing Nicolas's
>> changes made in the meantime... 
> 
> Version 8 removes roerf finally. :-) But I could not find any trace of
> #nextHandlerContext in the current changeset, did you maybe forget to
> revert the previous version before loading v7?

yes indeed, I forgot to remove the previous version, sorry for confusion,
everything's fine :)


Christoph Thiede wrote
>> What would you think about this approach: because #return:from: supplies
>> the first unwind context for #aboutToReturn:through: prematurely, how
>> about to supply nil instead of the first unwind context and let
>> #resume:through: find the first unwind context at precisely the right
>> time?
> 
> Correct me if I'm wrong, but this only would move the problem again,
> wouldn't it? If you press over too late, we would have the same problem
> again? I'd still prefer a holistic approach such as my
> #informDebuggerAboutContextSwitchTo: proposal. Or did miss anything
> different with your proposal? :-)

Well, I thought the fix really solved the bug, not just pushed it further
away :) I couldn't reproduce the incorrect stepOver behavior any longer but
I may have missed some example - do you have something in mind? I'll comment
further in [1]; I hope I'm not wrong here - please send a counterexample if
you find one :)

best,
Jaromir


Christoph Thiede wrote
> Best,
> Christoph

[1]
http://forum.world.st/BUG-REGRESSION-while-debugging-Generator-gt-gt-nextPut-td5108125i20.html



-----
^[^ Jaromir
--
Sent from: http://forum.world.st/Squeak-Dev-f45488.html