What is interesting is to be able to reimplement #ifTrue:ifFalse: and co in any object in the system and to use it. Now in practice it is difficult performance wise. As an experiment we overrided mustBeBooleanIn: so #ifTrue:ifFalse: and co would work on any object. However this hack is slow and interesting only for experiments IMO. I asked in the bug report to add a setting... But no one answered and it was integrated. Marcus said we'll add a setting later. 2014-08-06 8:31 GMT+02:00 stepharo <stepharo@free.fr>:

Ok thanks. I do not see why this is interesting to have that.

https://pharo.fogbugz.com/default.asp?13779

On 5/8/14 21:24, Clément Bera wrote:

Because someone activated the method "mustBeBooleanMagicIn:" by default whereas it was just a hack to experiment.

Basically you have an optimized message sent to a non boolean. For over a week now, in this case the code is executed as a DoIt with a non optimized message instead of raising the mustBeBoolean error.

2014-08-05 18:40 GMT+02:00 stepharo <stepharo@free.fr>:

...

Hi guys

Does anybody have an idea why when I open a transcript and recompile a method I get

UndefinedObject>>ExecuteUnOptimizedIn: (model is Undeclared)

UndefinedObject>>ExecuteUnOptimizedIn: (model is Undeclared)

UndefinedObject>>ExecuteUnOptimizedIn: (scroller is Undeclared)

Stef

On 5 août 2014, at 21:24, Clément Bera <bera.clement@gmail.com> wrote:

Because someone activated the method "mustBeBooleanMagicIn:" by default whereas it was just a hack to experiment.

If it's not integrated it's not gonna improve. This situation is a proof of that: it's because it's activated that we discover a bug (that is not the real problem BTW): wrong compilation context because of a line I kept from the old version of mustBeBooleanInMagic: (https://pharo.fogbugz.com/f/cases/13780/Wrong-compilation-context-in-mustBeB...). If you feel uneasy about that hack, please make a preference. So in the end I dont't understand the problem nor your tone...

Basically you have an optimized message sent to a non boolean. For over a week now, in this case the code is executed as a DoIt with a non optimized message instead of raising the mustBeBoolean error.

And instead you get a DNU, what's wrong with that? Why is mustBeBoolean any better than a DNU? For me the problem is to have ifTrue: sent to a PackageTreeNautilusUI inside a "on: Error do: [ "nothing" ]" (in PluggableTextMorph>>acceptBasic). So if anything, this hack acted as a problem revealer at the very worst.

2014-08-05 18:40 GMT+02:00 stepharo <stepharo@free.fr>: Hi guys

Does anybody have an idea why when I open a transcript and recompile a method I get

UndefinedObject>>ExecuteUnOptimizedIn: (model is Undeclared)

UndefinedObject>>ExecuteUnOptimizedIn: (model is Undeclared)

UndefinedObject>>ExecuteUnOptimizedIn: (scroller is Undeclared)

Stef