On Wednesday, 31 August 2011, Stéphane Ducasse <stephane.ducasse@inria.fr> wrote:
I think it is a very bad idea to integrate the refactoring engine into the system, but I guess we don't have to discuss that again.
This is simple. If nobody maintains the Configuration then it means that this is impossible to load it automatically We do not want a script we want a metacello spec for obvious reason. And it means that we are always confused and we do not know what is the latest version and on which repository it is.
I never said Metacello configs are evil. A configuration management is definitely needed. It is just too complex for me, given that I don't gain anything myself.
Then we do not want it to take 20 min to load because in that case it means that we will never be able to load it to integrate and work on core. Because when we work on core and integrate and fix things we generate images sometimes every 5 minutes so a jenkins server does not help at all.
On my dead slow single-core server from the year 2000 it takes 5 minutes to load RB, OB and run all test and lint rules (if no other builds run at the same time). I assume that it is possible to load the necessary packages on a modern machine in under 1 min right after each commit. Also note that with builder.sh one can do incremental builds, that is to update the image from the previous build. I am not currently using this trick anywhere, but it makes builds even on my slow machines almost instantanious. Lukas -- Lukas Renggli www.lukas-renggli.ch