Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
[incubation] continuous integration of another Eclipse project...

The OpenJ9 project proposal is in its creation review at this point (https://projects.eclipse.org/proposals/eclipse-openj9). This project will be consuming the Eclipse OMR project on a regular (actually, continuous, is what we would like) basis. These two projects are quite closely linked (in that Eclipse OMR provides Virtual Machine technology components, like garbage collection and JIT compilers, that are then "specialized" by OpenJ9 to implement a Java Virtual Machine).

What we're imagining is that OpenJ9 will maintain its own private "fork" of Eclipse OMR (an openj9.omr repository alongside the primary openj9 repository we're optimistically expecting to be created at GitHub) that we want to mirror commits from the upstream Eclipse OMR project when they happen. We really do think we want to be at the bleeding edge on both projects at the same time (because life isn't challenging enough, I guess). One motivating reason for that is so that OpenJ9 can immediately test and react to the impact of Eclipse OMR commits in the context of a Java Virtual Machine (i.e. leveraging tests written in Java).

I'm wondering primarily how this kind of consumption model should interact with the CQs one normally files when the "version" of a dependent project is being updated: if OpenJ9 pulls in every Eclipse OMR commit, which could mean up to 40+ commits per week, I'm *really* *really* hoping that will not require a CQ each. Another possibly relevant wrinkle is that, despite lots of activity, Eclipse OMR hasn't done a release yet (yeah, we're proud :( ).

Does any other project do this kind of thing already?

Any advice or, especially, feedback on how we can pragmatically use this kind of model and meet our IP tracking requirements?

Is there anything Eclipse OMR needs to do to facilitate this consumption model, and is there anything we need to ensure in the construction of OpenJ9 (or its private fork of OMR) to make this work and be practical?

--mark


Back to the top