Fixed OSGI import version for Grizzly#6050
Conversation
|
@senivam Could you do something with branch names? It is a mess, I am not sure which branch should be the target one. Why so old branch is the default one? Should I change it to 4.0 or 4.0.0-BRANCH is the right one? Why it declares 4.0.99-SNAPSHOT version as the next snapshot? I think it would be better to follow some Maven conventions.
|
|
Branch 4.0.0-Branch should be deleted, it is the release branch, from which is tag 4.0.0 created, I think. |
This doesn't restrict it to 3, but the meaning is 3 or later. Note that AI gives wrong answers, surprisingly SO is correct, see: https://stackoverflow.com/questions/8353771/osgi-valid-version-ranges/8354987#8354987 See also eclipse-ee4j/glassfish#25859 - just because Grizzly releases 5, we have to release some 5 projects despite they use still the same API. With Grizzly 6 it would repeat again for no reason. |
|
Sounds ok, then. Can you rebase onto a different branch? |
- We use 4.0.2 for build - The range was limiting major upgrades of other projects while breaking changes in Grizzly still did not affect Jersey. Now we set just the minimal Grizzly version. Signed-off-by: David Matějček <david.matejcek@omnifish.ee>
Done, I hope this branch is the right one :-) |
|
@dmatej - {version}-BRANCH branches are technical branches which are being created during a release process of a particular {version}. Usually they are merged back to the branches from which the release had occured and are being deleted aftewards. Some. hovever, are not merged yet that's why you are facing the |
Do we need the separate 4.0 branch still? Maybe just continue with 4.0.JPMS exclusively? |
@arjantijms - yes, good point, it was planned as you are saying - convert 4.0.JPMS into the 4.0 and by that replace the original 4.0 without JPMS. However, we are not 100% sure JPMS is totally fine, so that is why the original 4.0 branch is being kept. |
|
@senivam I renamed 4.0 to backup40branch, and renamed 4.0.JPMS to 4.x |

Note: I have noticed some issues with test stability (already without this change), I will try to improve it in standalone PR(s).