On Feb 1, 2011, at 8:58 37AM, Nicolas Cellier wrote:

2011/2/1 Marcus Denker <marcus.denker@inria.fr>:

On Feb 1, 2011, at 8:01 AM, Nicolas Cellier wrote:

2011/2/1 Marcus Denker <marcus.denker@inria.fr>:

Maybe I had some bad luck with
http://code.google.com/p/pharo/issues/detail?id=3468
Or is there many examples of issues closed without a clue ?

In a case like that, the changeset was just merged in the release.

The test is failing, as you can see:

https://pharo-ic.lille.inria.fr/hudson/job/Pharo%20Core%201.3/lastCompletedBuild/testReport/KernelTests.Methods/CompiledMethodTest/testPerformInSuperclassCanExecutelongMethodWithTemps/



This is probably because you must
   (Object recompile: #perform:withArguments:)


The 1.3 core image was recompiled completely around 13020...

       Marcus


--
Marcus Denker  -- http://www.marcusdenker.de
INRIA Lille -- Nord Europe. Team RMoD.



Strange, the change worked for me in Squeak once I recompiled.

Nicolas


#perform:withArguments: works in Pharo as well, Igor's fix for it was applied back in May:
http://code.google.com/p/pharo/issues/detail?id=2443

What it doesn't fix (and which the test in 3468 highlights), are arbitrary sends to 
#perform: withArguments: inSuperclass: (Ie. the compiler does not expand frame size of all senders, nor does the test run on a VM which does not error with many arguments)

#perform:withArguments: just happend to be the most likely sender to run into it.

Cheers,
Henry