Ø
The history start again : We have already faced such situation few years again. Some architectural resolutions and rules have been taken, but it seems that we have slipped again :-).
Not exactly: the refactored plug-ins never completely matched the defined architecture, so we didn’t really “slip again”
:)
We’ve made a first step in Papyrus 0.9.0/Juno, and now we’re trying to finish the work (But realistically, this will probably be “another step” rather
than “the last step”, because Neon M6 is not so far away).
Camille
De : mdt-papyrus.dev-bounces@xxxxxxxxxxx [mailto:mdt-papyrus.dev-bounces@xxxxxxxxxxx]
De la part de Cedric Dumoulin
Envoyé : vendredi 15 janvier 2016 14:03
À : Papyrus Project list <mdt-papyrus.dev@xxxxxxxxxxx>
Objet : Re: [mdt-papyrus.dev] Papyrus Bundle Dependency Refactorings for Neon
Hi,
The history start again : We have already faced such situation few years again. Some architectural resolutions and rules have been taken, but it seems that we have slipped again :-).
It looks like a lot of problems have been identified, and some actions are planned.
Is there somewhere some good UML diagrams of the Papyrus projects and their dependencies ? This could help to understand what actually goes wrong, and what should be done. This will also help in future.
cheers,
Cedric
Christian W. Damus a écrit :
Hi, Team,
I have begun an analysis of bundle dependency refactorings to bring the Papyrus architecture into shape at that level of granularity in the Neon release. The principal emphases of this effort are:
·
splitting UI dependencies out of core bundles that should provide their APIs to “headless” clients, for example in automation or web-service use cases
·
eliminating dependencies “pointing up” the layer stack, such as infra bundles’ usage of UML bundles, usage of GEF 3 in diagram framework bundles that should also support GEF 4, etc.
·
realizing the componentization of Papyrus so that build and test execution may be implemented on a finer granularity (Hello, test execution in Gerrit)
This will necessarily be a long and probably very disruptive process, with APIs in use all over the place moving hither and yon or simply disappearing. The dependency analysis is being tracked
·
in
bug 485220
·
on the
Papyrus wiki
·
in initial refactoring changes for UI dependencies in the Infra Layer:
change 63777 and
change 64168
·
there will be further Gerrit patches as work progresses, always linked from the bugzilla, of course
If and when these patches start to be merged, you will find the ground shifting under your feet as merge conflicts arise and APIs that you were using in new code no longer resolve.
It is not yet clear if and when any of these changes will be merged. I shall send another notice on the mailing list when the first change is merged, which will then effectively commit us to a Papyrus 2.0. At that time also a Migration Guide
will be published to cover those changes. The rest of the process will then just continue as it happens, with piecemeal refactoring patches and migration guide updates from time to time.
Cheers,
Christian
