On Mar 4, 2010, at 11:38 AM, Henrik Johansen wrote:
> FWIW, this quote from the following discussion on Squeak-dev distills what I was trying (but failing) to demonstrate:
> "We shouldn't stop trying to support well-behaved ensure blocks properly.
> Because if we don't then ensure
> becomes entirely random. Or can anyone here explain why it would be
> reasonable to expect that this version works:
>
> p := [[Processor yield] ensure:[Transcript show: 'finished']] fork.
> Processor yield.
> p terminate.
>
> but this one doesn't:
>
> p := [[] ensure:[Processor yield. Transcript show: 'finished']] fork.
> Processor yield.
> p terminate.
>
> (we're using yield here to denote any kind of process switch) If the
> only thing that determines whether an ensure block is executed is a
> process switch, then we basically have random, non-predictable unwind
> handling. Which is a problem that is entirely unrelated to whether the
> unwind block is itself well-behaved or not. "
>
> In a book with chapters discussing topics such as ensure and processes, (based on Pharo 1.0 which is unlikely to contain any fixes for such an issue), it deserves at least a mention.
>
> Cheers,
> Henry
>
> Den 03.03.2010 15:33, skrev Levente Uzonyi:
>> On Wed, 3 Mar 2010, St�phane Ducasse wrote:
>>
>>>
>>> On Mar 2, 2010, at 10:33 PM, Levente Uzonyi wrote:
>>>
>>>> On Tue, 2 Mar 2010, Henrik Sperre Johansen wrote:
>>>>
>>>>>
>>>>>>> PS. For the not-so-faint-of-heart, open a Transcript and try evaluating:
>>>>>>> |proc|
>>>>>>> proc := [[Transcript show: 'Start!'. ] ensure: [Transcript show:
>>>>>>> 'Finish!' Processor yield.]] newProcess.
>>>>>>> proc resume.
>>>>>>> Processor yield.
>>>>>>> proc suspend.
>>>>>>> proc terminate.
>>>>>> Why don't we get it printing?
>>>>> Forgot a . there, supposed to be
>>>>>
>>>>> |proc|
>>>>> proc := [[Transcript show: 'Start!'. ] ensure: [Transcript show:
>>>>> 'Finish!'. Processor yield.]] newProcess.
>>>>> proc resume.
>>>>> Processor yield.
>>>>> proc suspend.
>>>>> proc terminate.
>>>>>
>>>>> on my machine it prints:
>>>>> *Start!Start!Finish!
>>>>
>>>> The problem occurs because Transcript >> #endEntry (sent from #show:)
>>>> takes a while to execute, so the process (proc) prints 'Start!', but
>>>> it gets terminated before execution reaches #resetContents (#reset in
>>>> Squeak). So 'Start!' is still in the stream. Then our process executes the
>>>> #ensure: block and it prints 'Start!' and 'Finish!' too.
>>>
>>>
>>> Thanks for the explanation levente!
>>> But I'm still dizzy :) why Transcript show: 'Finish' would also print 'start' ?
>>
>> Transcript is just a stream. #show: is just a combination of #nextPutAll: and #endEntry. #endEntry displays the contents of the stream and when that's done it removes everything from the stream with #resetContents. #resetContents is never executed, because the process is terminated before that. So the contents of the stream are kept and displayed again later.
>>
>>
>> Levente
>>
>>>
>>>>
>>>> There's worse problem with #ensure: and #terminate is that, a process executing an #ensure: block can be terminated. If you evaluate this code:
>>>>
>>>> | process stage1 stage2 stage3 counter |
>>>> stage1 := stage2 := stage3 := false.
>>>> counter := 0.
>>>> process := [
>>>> � � [ stage1 := true ] ensure: [
>>>> � � � � stage2 := true.
>>>> � � � � 1000000 timesRepeat: [ counter := counter + 1 ].
>>>> � � � � stage3 := true ] ] newProcess.
>>>> process resume.
>>>> Processor yield.
>>>> process suspend.
>>>> process terminate.
>>>> 1000 milliSeconds asDelay wait.
>>>> { stage1. stage2. stage3. counter } explore
>>>>
>>>> you will find that stage1 and stage2 is reached as expected, but stage3 is not and counter is less than 1000000. That's because the forked process started evaluating the #ensure: block, but it was terminated by our process.
>>>>
>>>> Is this the expected behavior when sending #terminate to a process which is evaluating an #ensure: block?
>>>>
>>>>
>>>> Cheers,
>>>> Bal�zs & Levente
>>>>
>>>>
>>>>>
>>>>> *yay
>>>>> Henry*
>>>>> *
>>>>
>>>
>>>
>>> _______________________________________________
>>> Pharo-project mailing list
>>>
Pharo-project@lists.gforge.inria.fr
>>>
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>>>
>>
>> _______________________________________________
>> Pharo-project mailing list
>>
>>
Pharo-project@lists.gforge.inria.fr
>>
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project