W3C home > Mailing lists > Public > www-voice@w3.org > January to March 2013

Re: Bug with deep and shallow history in pseudo-code

From: chris nuernberger <cnuernber@gmail.com>
Date: Sat, 9 Feb 2013 16:04:51 -0700
Message-ID: <CAG=GWvdqbyS0ayDSAUoPSq7zQENqQGsePjQZDTx80WPnnEYmmQ@mail.gmail.com>
To: Jim Barnett <Jim.Barnett@genesyslab.com>
Cc: Stefan Radomski <radomski@tk.informatik.tu-darmstadt.de>, "www-voice@w3.org (www-voice@w3.org)" <www-voice@w3.org>
In this case that isn't true.

The parallel state is in the target list of the transition fired by t6.
 The parallel states first child, its history is the history in question.
 addStatesToEnter will be called with the parallel state as its first
argument.  It will fall down into the second clause (because the parallel
state isn't a history state) that reads:

statesToEnter.add(state)
        if isCompoundState(state):
            statesForDefaultEntry.add(state)
            for s in getTargetStates(state.initial):
                addStatesToEnter(s,statesToEnter,statesForDefaultEntry)
        elif isParallelState(state):
            for s in getChildStates(state):
                addStatesToEnter(s,statesToEnter,statesForDefaultEntry)



Finding the elif because state is indeed a parallel state.

If getChildStates returns p's history state (which is its first state) then
you get the problem Stefan described.

If it does not (which I think we both agree that it should not) then you
get different behavior and the problem Stefan is concerned about doesn't
happen.

Chris

On Sat, Feb 9, 2013 at 3:51 PM, Jim Barnett <Jim.Barnett@genesyslab.com>wrote:

>  No, but by the time this is called, the history state would have been
> de-referenced, and the actual history values would be on the statesToEnter
> list.****
>
> ** **
>
> **-          **Jim****
>
> ** **
>
> *From:* chris nuernberger [mailto:cnuernber@gmail.com]
> *Sent:* Saturday, February 09, 2013 5:46 PM
> *To:* Jim Barnett
> *Cc:* Stefan Radomski; www-voice@w3.org (www-voice@w3.org)
> *Subject:* Re: Bug with deep and shallow history in pseudo-code****
>
> ** **
>
> getChildStates would not return a history state, would it?****
>
> ** **
>
> Chris****
>
> ** **
>
> On Sat, Feb 9, 2013 at 3:22 PM, Jim Barnett <Jim.Barnett@genesyslab.com>
> wrote:****
>
> Stefan,
>   This clause in enterStates is supposed to catch that case:
>
>                     statesToEnter.add(anc)
>                     if isParallelState(anc):
>                         for child in getChildStates(anc):
>                             if not statesToEnter.some(lambda s:
> isDescendant(s,child)):
>
> addStatesToEnter(child,statesToEnter,statesForDefaultEntry)
>
> but it may be missing something.  I need to look at enterStates and
> addStatesToEnter more closely.  At the very least, they need to be
> reorganized for clarity.
>
> = Jim****
>
>
> -----Original Message-----
> From: Stefan Radomski [mailto:radomski@tk.informatik.tu-darmstadt.de]
> Sent: Saturday, February 09, 2013 1:04 PM
> To: www-voice@w3.org (www-voice@w3.org)
> Subject: Bug with deep and shallow history in pseudo-code
>
> Hi again,
>
> the SCXML document at [1] will put my interpreter in an illegal state
> configuration. I am sure that there is a more concise example that would
> accomplish the same, but it is part of Jacob's test framework.
>
> Initial:
>   Enabled transitions: {root -> a}
>   Stable initial configuration: {a}
>
> t1 arrives:
>   Enabled transitions: {a -> p}
>   Stable configuration: {p, b, b1, b1.1, c, c1, c1.1}
>
> t2 arrives:
>   Enabled transitions: {b1.1 -> b1.2, c1.1 -> c1.2}
>   Stable configuration: {p, b, b1, c, c1, b1.2, c1.2}
>
> t3 arrives:
>   Enabled transitions: {b1.2 -> b2, c1.2 -> c2}
>   Stable configuration: {p, b, c, b2, b2.1, c2, c2.1}
>
> t4 arrives:
>   Enabled transitions: {b2.1 -> b2.2, c2.1 -> c2.2}
>   Stable configuration: {p, b, c, b2, c2, b2.2, c2.2}
>
> When t5 arrives, we exit everything and enter "a" again, thus we have to
> save three histories:
>   hp: {b2.2, c2.2}
>   hb: {b2.2}
>   hc: {c2}
>   Stable configuration: {a}
>
> t6 will transition into "p" again and bring the interpreter into an
> illegal configuration:
>   addStatesToEnter will realize that "p" is parallel an attempt to enter
> all its children
>     Entering hp will enter {b2.2, c2.2} as its history
>     As "p" is parallel, we will enter "c" as well, as it is compound, we
> will enter its initial shallow history "hc", adding "c2"
>       "c2" is compound and its initial state "c2.1" will be entered.
>
> -> we now have "c2.2" and "c2.1" in the configuration which is illegal.
>
> Who is supposed to prevent "c2" from entering its initial state? We
> already entered "c2.2" per deep history from "hp" and now the shallow
> history from "hc" wants to enter "c2" and ultimately its initial state
> "c2.1" leading to an invalid configuration.
>
> Best regards
> Stefan
>
> [1]
> https://github.com/jbeard4/scxml-test-framework/blob/master/test/history/history4.scxml
>
>
>
>
> ****
>
>
>
> ****
>
> ** **
>
> --
> A foolish consistency is the hobgoblin of little minds - Emerson ****
>



-- 
A foolish consistency is the hobgoblin of little minds - Emerson
Received on Saturday, 9 February 2013 23:05:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Saturday, 9 February 2013 23:05:22 GMT