Stef wrote: > Doru wrote: >> We need to reconsider how we approach continuous integration for >> pharo-based images. >can you expand ?
The most important thing, from my POV, is to not break things silently when cleaning up the system. Having a system integration test that tells you you are no longer supposed to have >>packages methods, makes things easier to fix. Should we tag tests with <integration> to make clear what tests are expected to be run in the 'leaf' builds? Marcus wrote: >But *Grease* still is the problem: in the tradition of the wonderful >philosophy “Lets keep all smalltalk crappy >and provide some layer on top”, it redefines #package and therefore requires >that no Smalltalk ever >implements #packages on the class side to return the system concept of >package. That's not a really helpful remark. The purpose of Grease is to provide a compatibility layer to let a.o. Seaside and Magritte work on as many Smalltalks as possible. In 2009, using a class-side packages method in GRPackage was not a bad idea, AFAIK. The name was not claimed, and later used by several other classes. By not having a test claiming the use of >>packages for RPackage, (and you might ask, why not rPackages?), and having lots of old flag: #todo code in RPackage, it is not easy for Grease maintainers to find out that that renaming is the obvious thing to do. It is part of the Grease API, and making changes there always involves checking all platforms. AFAIK, no maintainer has regular access to all platforms. Stephan
