Hi Tas,
Tas Frangoullides schrieb:
This does sounds interesting. Does this mean that the target platform
would not have to be re-created everyime? That could speed up the builds
considerably. It could be esepcially useful where a few products use the
same target platform. For example RCP+EMF is a common one for me.
Yes, that's exactly what I had in mind. Let's say a company provides one
or more target platforms for all the eclipse products. Seperating the
target platform creation from the actual product builds would increase
reusablitiy and the individual builds would IMO be simplified a lot (the
target platform is just there, you just need to select a proper one
instead of building it from scratch).
Do you expect to use a seperate hudson task for creating the target
platform? It would mean that your product build is structured well so
that hudson can bypass the target platform creation, which your command
line builds would not.
Yes, the target platform creation is a separate hudson task.
Why would I have command line builds and hudson builds at the same time?
There's hudson on the server side and there's the developers on the
'client' side. Developers have their own target platform anyway I'd say
and since products created locally on a developers machine would never be
released, there's no point in keeping the server TP and the local TP
perfectly in-sync I'd say (although a developer can easily download the TP
produced by hudson and set that as their target platform, of course).
For a developer it would simply be buckminster -> invoke action -> create
product and the same would happen in hudson. Only difference is the target
platform.
Best regards,
Johannes