| Summary: | [UML 2.5] CNFE for L2Package for UML 2.5 | ||
|---|---|---|---|
| Product: | [Modeling] Papyrus | Reporter: | Ed Willink <ed> |
| Component: | Core | Assignee: | Camille Letavernier <cletavernier> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | P3 | CC: | cletavernier |
| Version: | 1.0.0 | ||
| Target Milestone: | M6 | ||
| Hardware: | All | ||
| OS: | All | ||
| Whiteboard: | |||
| Bug Depends on: | 429744 | ||
| Bug Blocks: | |||
|
Description
Ed Willink
> This should not occur because Papyrus N-builds should be using UML I-builds and so should detect the problem.
No; Papyrus uses the latest builds from Luna/Staging. We want our nightly builds to be installed on the latest Eclipse Milestones.
We have too many dependencies; we can't ask our early adopters to use all possible I-build update sites (Moreover, most I-build update sites are not synchronized with each other, whereas Luna/staging is; that's the point of simultaneous release).
We will migrate to UML2.5 a few days before M6.
It depends on who the builds are for. In the early Xtext days I got fed up with discovering a new build nightmare two days before a milestone by which time it was too late to get the API change reverted. Doing I-builds against I-builds and N-builds against N-builds ensures that evolutions are resolved as soon as possible. Before M6, we anticipate a few more days to fix everything. That's where the APIs change the most. But we cannot afford this extra-work for each and every milestones, I-build and nightly build. We have way too many dependencies, and 95% of the time, 3 days are enough to catch up. For the remaining 5%, it's easier to simply disable some features for the incoming milestone (And take a few more days when M6 is approaching). (BTW, our nightly builds target early adopters in priority, which means they have to be usable as much as possible, without too many installation issues) The migration to UML 2.5 is complete (Luna M6). I close the task |