Fix the other dependencies of asContextWithSender:

Eliot

should this fix be applied to pharo?
       http://code.google.com/p/pharo/issues/detail?id=2655

Is there a dependency?

Stef

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Hi Eliot,

There are several other changes in that package that could generate different behavior. Are they correct? They are:

BlockClosure >> value
  used to call primitive 201, now calls 205
  used to check for zero args, now checks for one (!?)

BlockClosure >> value:
  used to call primitive 202, now calls 205

BlockClosure >> value:value:
  used to call primitive 203, now calls 205

BlockClosure >> value:value:value:
  used to call primitive 204, now calls 205

Primitive 205 was previously only used for #value:value:value:value:, so I wonder if this might be a mistake.

Thanks,
Juan Vuletich

Eliot Miranda wrote:

On Tue, Jul 6, 2010 at 12:51 PM, stephane ducasse <[hidden email] <mailto:[hidden email]>> wrote:

   Eliot

   should this fix be applied to pharo?
          http://code.google.com/p/pharo/issues/detail?id=2655


Yes please.  Almost everything I'm likely to put in Compiler, Kernel or System on Squeak 4.1 belongs in all dialects using the closure compiler.
 

   Is there a dependency?


What's happing here is that very old primitive failure code that allowed the closure system to run on old VMs without closure support depended on BlockClosure>>asContext, which had a bug in it.  So the changes reflect the bug fix but are enclosed in false ifTrue: [] because they really are obsolete, but the code is retained because its useful documentation of how the closure primitives should work.  I hope this makes sense!

best
Eliot
 

   Stef

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project

Hi Eliot,

Eliot Miranda wrote:

Hi Juan,

On Wed, Jul 7, 2010 at 5:24 AM, Juan Vuletich <[hidden email] <mailto:[hidden email]>> wrote:

   Hi Eliot,

   There are several other changes in that package that could
   generate different behavior. Are they correct? They are:

   BlockClosure >> value
     used to call primitive 201, now calls 205
     used to check for zero args, now checks for one (!?)

   BlockClosure >> value:
     used to call primitive 202, now calls 205

   BlockClosure >> value:value:
     used to call primitive 203, now calls 205

   BlockClosure >> value:value:value:
     used to call primitive 204, now calls 205

   Primitive 205 was previously only used for
   #value:value:value:value:, so I wonder if this might be a mistake.


Quite right, this is a mistake.  I was careless copy-pasting the fix between them.  I've committed fixes to trunk.  In the VM these primitive numbers all bind to the same primitive so there was no visible effect.  Thanks for checking!

best
Eliot

When I integrate stuff in Cuis, I always do my best effort to understand it!

Anyway, I see your new version, but in BlockClosure >> value it still does
  numArgs ~= 1 ifTrue: [self numArgsError: 1].
even if the older version (9/3/2008) it checked for zero arguments. Is this right? If so, what's the rationale?

Thanks!
Juan Vuletich

_______________________________________________
Pharo-project mailing list
[hidden email]
http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/pharo-project