Fwd: Re: FFI in closure image/modifying block arguments

> FYI: with the new closure changes/image (Pharo0.1Core-10268cl.zip from [1])
> it is not possible to store into block arguments:
>
>   1 to: 10 do: [:i | i := i + 1. Transcript show: i printString ]
>
> That makes the current version of FFI unloadable, due to
> the code in  ExternalTypeClass>>initializeAtomicTypes
>
> Bye
> T.
>
>
> [1] https://gforge.inria.fr/frs/?group_id=1299

> I forwarded the mail from Andreas, since I think he is not subscribed
> to pharo list and his answer didnt appear on the the list archive.
>
> Many places - I agree. I had the problem while loading OB alpha before
> FFI since OB also had block arg assignment. But it was very easy to
> fix. FFI could/should also be fixed. That's at least a start.
>
> However - the question left is if block arg assignment should be
> allowed ...
>
> Bye
> T.
>
> BTW: Andreas: would be good if you could subscribe to pharo list. If
>     you fear the traffic then just disable mail forwarding and use
>     the web archive to read. That's how I do it on all squeak related
>     lists and it works fine.
>
>
> -------- Original-Nachricht --------
> Datum: Tue, 07 Apr 2009 07:06:41 -0700
> Von: Andreas Raab <[hidden email]>
> An: Torsten Bergmann <[hidden email]>
> CC: [hidden email]
> Betreff: Re: FFI in closure image/modifying block arguments
>
> I think it is advisable to allow block arg assignment for the time
> being. There is *lots* of code which uses it knowingly or unknowingly
> and unless you've fixed every single case you'll have these troubles
> every step of the way. For laughs, just try to recompile all the  
> classes
> in your image to see how far you get.
>
> Cheers,
>   - Andreas
>
> Torsten Bergmann wrote:
>> FYI: with the new closure changes/image (Pharo0.1Core-10268cl.zip  
>> from [1])
>> it is not possible to store into block arguments:
>>
>>  1 to: 10 do: [:i | i := i + 1. Transcript show: i printString ]
>>
>> That makes the current version of FFI unloadable, due to
>> the code in  ExternalTypeClass>>initializeAtomicTypes
>>
>> Bye
>> T.
>>
>>
>> [1] https://gforge.inria.fr/frs/?group_id=1299
>
> --
> Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit  
> allen: http://www.gmx.net/de/go/multimessenger01
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>

> I forwarded the mail from Andreas, since I think he is not subscribed
> to pharo list and his answer didnt appear on the the list archive.
>
> Many places - I agree. I had the problem while loading OB alpha before
> FFI since OB also had block arg assignment. But it was very easy to
> fix. FFI could/should also be fixed. That's at least a start.
>
> However - the question left is if block arg assignment should be
> allowed ...
>
> Bye
> T.
>
> BTW: Andreas: would be good if you could subscribe to pharo list. If
>     you fear the traffic then just disable mail forwarding and use
>     the web archive to read. That's how I do it on all squeak related
>     lists and it works fine.
>
>
> -------- Original-Nachricht --------
> Datum: Tue, 07 Apr 2009 07:06:41 -0700
> Von: Andreas Raab <[hidden email]>
> An: Torsten Bergmann <[hidden email]>
> CC: [hidden email]
> Betreff: Re: FFI in closure image/modifying block arguments
>
> I think it is advisable to allow block arg assignment for the time
> being. There is *lots* of code which uses it knowingly or unknowingly
> and unless you've fixed every single case you'll have these troubles
> every step of the way. For laughs, just try to recompile all the  
> classes
> in your image to see how far you get.
>
> Cheers,
>   - Andreas
>
> Torsten Bergmann wrote:
>> FYI: with the new closure changes/image (Pharo0.1Core-10268cl.zip  
>> from [1])
>> it is not possible to store into block arguments:
>>
>>  1 to: 10 do: [:i | i := i + 1. Transcript show: i printString ]
>>
>> That makes the current version of FFI unloadable, due to
>> the code in  ExternalTypeClass>>initializeAtomicTypes
>>
>> Bye
>> T.
>>
>>
>> [1] https://gforge.inria.fr/frs/?group_id=1299
>
> --
> Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit  
> allen: http://www.gmx.net/de/go/multimessenger01
>
> _______________________________________________
> Pharo-project mailing list
> [hidden email]
> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>

> Stéphane Ducasse wrote:
>> block args in smalltalk are read-only
>
> For sure. But my point is that there is a lot more code out there  
> using block arg assignment than you might think. And it's one issue  
> to keep your own code clean, and a completely different one to force  
> other people to change theirs to your standards. Most of what you'll  
> get is complaints that "your system does't work, it doesn't even  
> load code that works fine in x, y, or z". Been there, done that  
> (underscore assignment in Croquet). It's a difficult road since you  
> loose potential users (people try, fail, and go away). But do  
> whatever you feel like.
>
> Cheers,
>  - Andreas
>
>
>> On Apr 7, 2009, at 10:01 PM, Torsten Bergmann wrote:
>>> I forwarded the mail from Andreas, since I think he is not  
>>> subscribed
>>> to pharo list and his answer didnt appear on the the list archive.
>>>
>>> Many places - I agree. I had the problem while loading OB alpha  
>>> before
>>> FFI since OB also had block arg assignment. But it was very easy to
>>> fix. FFI could/should also be fixed. That's at least a start.
>>>
>>> However - the question left is if block arg assignment should be
>>> allowed ...
>>>
>>> Bye
>>> T.
>>>
>>> BTW: Andreas: would be good if you could subscribe to pharo list. If
>>>    you fear the traffic then just disable mail forwarding and use
>>>    the web archive to read. That's how I do it on all squeak related
>>>    lists and it works fine.
>>>
>>>
>>> -------- Original-Nachricht --------
>>> Datum: Tue, 07 Apr 2009 07:06:41 -0700
>>> Von: Andreas Raab <[hidden email]>
>>> An: Torsten Bergmann <[hidden email]>
>>> CC: [hidden email]
>>> Betreff: Re: FFI in closure image/modifying block arguments
>>>
>>> I think it is advisable to allow block arg assignment for the time
>>> being. There is *lots* of code which uses it knowingly or  
>>> unknowingly
>>> and unless you've fixed every single case you'll have these troubles
>>> every step of the way. For laughs, just try to recompile all the  
>>> classes
>>> in your image to see how far you get.
>>>
>>> Cheers,
>>>  - Andreas
>>>
>>> Torsten Bergmann wrote:
>>>> FYI: with the new closure changes/image (Pharo0.1Core-10268cl.zip  
>>>> from [1])
>>>> it is not possible to store into block arguments:
>>>>
>>>> 1 to: 10 do: [:i | i := i + 1. Transcript show: i printString ]
>>>>
>>>> That makes the current version of FFI unloadable, due to
>>>> the code in  ExternalTypeClass>>initializeAtomicTypes
>>>>
>>>> Bye
>>>> T.
>>>>
>>>>
>>>> [1] https://gforge.inria.fr/frs/?group_id=1299
>>>
>>> --
>>> Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit  
>>> allen: http://www.gmx.net/de/go/multimessenger01
>>>
>>> _______________________________________________
>>> Pharo-project mailing list
>>> [hidden email]
>>> http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project
>>>
>