There's definitely some fishy interaction between some suspect method definitions in Coral and RPackage. I need to check that I am actually using Camillo's fixes for that. In an image without RPackage, things go a little better (I can proceed through the loading, then some code is broken because of missing classes, but at least it loads without popping a hundred debuggers up). On 5 October 2011 18:49, Dale Henrichs <dhenrich@vmware.com> wrote:
Damien,
Some reproducible test cases would be handy here.
The behavior of Metacello is dependent upon the state of the image. Whether or not a package is loaded and which version of a package is loaded depends upon the starting state of an image ... if you load things by hand or have a partially loaded system, Metacello will do something different the next time a load expression is run...
Dale
----- Original Message ----- | From: "Damien Pollet" <damien.pollet@gmail.com> | To: Pharo-project@lists.gforge.inria.fr | Sent: Wednesday, October 5, 2011 6:40:44 AM | Subject: Re: [Pharo-project] RPackage / Metacello mystery woes | | â¦and of course it's a heisenbug so now it loads, provided I proceed | in | the monticello warning⦠| | On 5 October 2011 07:54, Stéphane Ducasse <stephane.ducasse@inria.fr> | wrote: | > OK how do I produce that? | > | > I do not understand why the method is sent multiple time.... | > | > Now I do not understand why the code of Coral refer to RPackage | > or petit* or .... | > | > In which version are you trying to load what? | > Stef | > | > | > On Oct 5, 2011, at 12:00 AM, Damien Pollet wrote: | > | >> On 4 October 2011 20:43, Stéphane Ducasse | >> <stephane.ducasse@inria.fr> wrote: | >>> I do not see why this would be related to RPackage. | >> | >> I don't know, that's just where things fail. The debuggers don't | >> help | >> much because by the time they display, the objects have changed | >> state⦠| >> | >> Camillo had a similar looking problem because of #package: being | >> defined in just the wrong place by Coral (for the class creation | >> convenience syntax) but he fixed that and there were working | >> builds | >> since then. | >> | >> | >>> On Oct 4, 2011, at 8:10 PM, Damien Pollet wrote: | >>> | >>>> Loading Coral is failing, apparently because something calls | >>>> #addMethod: on a Symbol, when it should have been an instance of | >>>> RPackage⦠| >>>> | >>>> See for instance | >>>> https://ci.lille.inria.fr/pharo/job/Coral/191/console | >>>> | >>>> I suspect it's coming from me updating ConfigurationOfCoral, | >>>> which | >>>> indirectly (via PetitParser) depends on | >>>> ConfigurationOfRefactoringBrowser. I've seen MC try to load | >>>> Refactoring-Tests-Core (lr.54) on top of (MarcusDenker.54), and | >>>> there | >>>> it complains some definitions are not there. If I proceed there, | >>>> code | >>>> seems to load but a bunch of debuggers open at the end, all with | >>>> a | >>>> similar error as the trace above. | >>>> | >>>> -- | >>>> Damien Pollet | >>>> type less, do more [ | ] http://people.untyped.org/damien.pollet | >>>> | >>> | >>> | >>> | >> | >> | >> | >> -- | >> Damien Pollet | >> type less, do more [ | ] http://people.untyped.org/damien.pollet | >> | > | > | > | | | | -- | Damien Pollet | type less, do more [ | ] http://people.untyped.org/damien.pollet | |
-- Damien Pollet type less, do more [ | ] http://people.untyped.org/damien.pollet