Re: test for default history content

On Tue, Apr 1, 2014 at 7:45 PM, Ate Douma <ate@douma.nu> wrote

>
> Sorry for chiming in this late, and going back to this earlier message,
> but I'm puzzled by the above statement.
>
> I can find no indication in the algorithm nor in the specification text
> that the getTargetStates function is supposed to do anything more than just
> returning the direct target(s) of a transition.
>
> It seems to me that the problems you encountered are actually *caused* by
> the the history-awareness and dereferencing in your getTargetStates
> implementation.
>
> AFAICT the pseudo algorithm expects it will take care of the history
> dereferencing itself in the addDescendantStatesToEnter procedure.
> If you do this upfront in you getTargetStates implementation, the
> addDescendantStatesToEnter never will get to 'see' a history state as it is
> supposed to, and neither can it then handle the default history content as
> just was added.
>
> I just completed the implementation of the current algorithm for Apache
> Commons SCXML (although not yet committed), and I now can successfully run
> test579, so it seems to me the proposed change simply works as expected.
>
> But maybe I'm overlooking some other specification requirements for
> getTargetStates?
> I'd appreciate some pointers in that case :)
>

Hi Ate,

Very glad you are chiming in.

As mentioned a number of times, my investigations around executable content
of history nodes, which you have brought to our attention, have been
limited to the surface and I'm simply debugging existing code to a large
degree.

It is true that my getTargetStates has history awareness ..likely as it
seemed necessary at the time of implementation.

If this is a mistake, or there exists an alternative, I'd be much obliged
to be enlightened.
The rest of my implementation follows the proposed spec virtually line for
line, so it would seem a little surprising, but it may be that code in some
other unspecified methods necessitate the history-awareness of my
getTargetStates method.

"Quite conveniently", I absolutely cannot spend more time on this for the
next few days again, so I hope you may arrive at better insights and
definitions with others here.

All the best,
Zjnue