Recently I took a look at the Global Keys project of Tony Fleig and noticed something strange. The package GlobalKeys makes some class extensions to the HandMorph class, extensions which are accessors to an instance variable. Since instance variable can't be included in class extensions, the variable appears as undeclared instead. HandMorph>>globalKeystrokeHandler "class extension" ^globalKeystrokeHandler "is an undeclared variable" HandMorph>>globalKeystrokeHandler: aOneArgBlock globalKeystrokeHandler := aOneArgBlock However, when loading the package, class side initialization uses the mutator #globalKeystrokeHandler: to set a default behavior for 'World activeHand' (an instance of HandMorph) and... it works! No error, it just works as if the instance variable is well defined and can be accessed without problem. GlobalKeys class>>setHandMorphHook "called by class initialization" World activeHand globalKeystrokeHandler: [ :ev | | retval | retval := false. "..." ] How can it be that the first invocation of #globalKeystrokeHandler: in #setHandMorphHook does not crash? Try the following: Gofer new squeaksource: 'GlobalKeys'; version: 'GlobalKeys-TonyFleig.9'; load Print it: World activeHand globalKeystrokeHandler ---> print the source of a code block Now debug the above snippet and proceed *into* globalKeystrokeHandler Print it again ----> print nil as if the code was recompiled once processed by the debugger You can also add the instance var globalKeystrokeHandler, recompile, it will be set to nil. -- View this message in context: http://forum.world.st/Weird-behavior-in-a-package-loaded-with-Monticello-tp3... Sent from the Pharo Smalltalk mailing list archive at Nabble.com.