I tried it locally in the recent PR, which upgraded to 2.13.15. Now, we need three configurations for all the cross-scala versions. It seems like Drop 2.12 and Java 8 are good options now.
I think there will be 100+ files that need to be changed. And because Scala 3.3.4 doesn't match the 2.13.15, I think we can't make it both works at the same time. 何品 Matthew de Detrich <mdedetr...@gmail.com> 于2024年10月22日周二 16:06写道： > > I would really like to avoid having to maintain several branches. > > I don't think there is a way around this if we want to respect users' > requirements and/or follow SemVer. Maintaining multiple branches is also > the norm when it comes to maintenance for non trivial sized projects and as > long as the projects are mainly source/feature compatible its not that much > more maintenance work, basically whenever a PR gets merged into 2.x.x we > will also cherry pick it in 1.x.x. > > That being said, we should pick an optimal time to do this i.e. when the > sbt build and other such features stabilizes for Pekko. I want to also > integrate some formatting rules into Pekko but I am waiting for a new > release of scalafmt, but basically we want to make sure as much as possible > that the build/formatting etc ete are "stable" and so most/all code drops > are just bug fixes/new features > > > So, I really hope that we can start doing 1.2 or 2.0 releases for some of > the modules soon. We already have some PRs that would ideally not appear in > a 1.1 release (new features, small API changes, dependencies upgrades that > break java 8 compat, etc.). > > Since we happen to be following strict semver, if we are going to do this > then it would need to be a v2.0. There are advantages to doing a 2 release > as since its a breaking release we can add features, i.e. > > * The inlining work for Scala 3 specifically (we had to roll this back in > Pekko 1.x because we accidentally broke bin compatibility and there was no > way around it) > * Remove all deprecated methods, this should really help with > maintenance burden > * Undo the @noinline changes specifically wrt to tracing/telemetry (we can > classify this as a breaking change) but also open up an official API with > opentelemetry/kamon > * Drop Java 8 support and have Java 11 as min > * Use Scala 3.6 LTS???? (really emphasize the ? here, I don't even know if > its a good idea but Scala 3 has solved a few issues in the next LTS that > was unsolvable in Scala 3.3 LTS series) > * Drop all akka <-> pekko migration features > * Upgrade to sbt 2.x for the build (this is coming out soon). > > The downside to this is that we need to maintain 2 branches and so do all > community plugins, but the pro's are also quite strong. We essentially can > reset from a clean slate and we only need to make sure that Pekko Cluster > upgrades work from 1.x to 2.x. Also if we plan to do this, release pekko > 2.x will take a bit longer but I think its worth it (there is no real rush > and if we are going to make a breaking version release we may as well do it > properly). > > > > On Wed, Oct 16, 2024 at 5:21 PM PJ Fanning <fannin...@apache.org> wrote: > > > I don't think we can support keeping all the version numbers in sync > > across all the Pekko modules. Pekko 1.1 is an exception because > > * we needed to roll out the Scala 2 inlining across all the modules > > * all our Pekko 1.0 modules are suffering from old dependencies due to > our > > decision to try to keep Pekko 1.0 dependencies as close as possible to > the > > last Akka Apache licensed releases to ease the switchover for Akka users > - > > and Pekko 1.1 modules have a much newer set of dependencies > > > > We can provide a doc page that lists our various modules and what > versions > > of other modules that they need. Also: as a place to keep track of the > > latest release numbers. > > > > I have a BOM project but that is of limited use because with sbt you need > > a plugin and some scripting to support BOMs. > > https://github.com/pjfanning/pekko-libraries-bom > > > > So, I really hope that we can start doing 1.2 or 2.0 releases for some of > > the modules soon. We already have some PRs that would ideally not appear > in > > a 1.1 release (new features, small API changes, dependencies upgrades > that > > break java 8 compat, etc.). > > > > I don't mind if we only remove Java 8 support in a few places. Pekko > > Connectors could become a mess though - with some connectors needing Java > > 11 or 17 minimum while others still support Java 8. Anything Slick > related > > will need Java 11 as HikariCP has driven them to now target Java 11. > > > > > > On 2024/10/14 10:33:26 Arnout Engelen wrote: > > > I would really like to avoid having to maintain several branches. > > > > > > I'd be in favour of dropping Java 8 support, but if that means we feel > > > pressure to maintain several branches, I'd rather merely officially > > > deprecate it. I feel similarly about Scala 2.12. Perhaps we can indeed > > > start by requiring a higher Java version in satellite projects, like > > > pekko-persistence-jdbc or pekko-connectors? > > > > > > I'm OK with dropping methods that were deprecated in Pekko 1.1.0 in > > > the next major/minor version. We can do that whether or not we go with > > > 1.2.0 or 2.0.0 for the version number if we follow > > > > > > https://pekko.apache.org/docs/pekko/current/common/binary-compatibility-rules.html > > > rather than 'strict semver'. > > > > > > We should decide on how synchronized version numbers across core and > > > satellite projects are. Ideally it should be easy to find out which > > > versions are compatible with each other. On the other hand, we should > > > be careful not to introduce too much churn on the maintainer side. > > > > > > If we release version 2.0.0 of pekko-management, that could still > > > depend on pekko-core 1.1.0, I think, right? But because there may be > > > breaking changes between pekko-core 1.x and 2.x, once we release > > > pekko-core 2.0.0, we should also release new major versions of all > > > satellite projects (i.e. 3.0.0 for pekko-management). So the invariant > > > is: "you must use a matching major version across transitive > > > dependencies, but you may upgrade to newer minor/patch versions of > > > transitive dependencies". > > > > > > This might be a reason to be sparse with major version updates, and at > > > least go with 1.2.0 for the next pekko-core version: this would save > > > us from having to do another round of releases of all satellite > > > projects that just bump versions. > > > > > > > > > Kind regards, > > > > > > Arnout > > > > > > On Fri, Aug 16, 2024 at 5:47 PM PJ Fanning <fannin...@apache.org> > wrote: > > > > > > > > We have another discussion open about doing a Pekko 1.1.0 release. > > > > > > > > After we get that released, we will have to do 1.1.0 releases for > > > > other other Pekko modules (HTTP, gRPC, Connectors, etc.). > > > > > > > > At the same time, we would need to decide on what to do about the > next > > > > release after that. > > > > > > > > I would suggest that we should consider making that next release a > > > > 2.0.0 release - one where we get to remove some deprecated code and > > > > potentially update the minimum Java and Scala versions that we > > > > support. > > > > We will continue to do patch releases for Pekko 1.0.x and Pekko 1.1.x > > > > so users who are affected by us dropping some things are not going to > > > > be too badly affected. > > > > > > > > * Drop Scala 2.12 support? The Scala 2.12 compiler has some type > > > > inference issues that complicate our code. The next Scala 3 LTS > > > > version looks like it will have some changes that will make it more > > > > different from Scala 2.12. > > > > * Go to Java 11 or even 17 as a minimum Java version? We already have > > > > issues in Pekko Connectors where we are stuck on older dependency > > > > versions because those dependencies have moved on from Java 8. > > > > * Drop the methods and classes that were deprecated in Akka before we > > > > split out Pekko? > > > > * Possibly remove some of the methods and classes that we deprecated > > in Pekko 1? > > > > > > > > What does everyone think? > > > > > > > > --------------------------------------------------------------------- > > > > To unsubscribe, e-mail: dev-unsubscr...@pekko.apache.org > > > > For additional commands, e-mail: dev-h...@pekko.apache.org > > > > > > > > > > > > > -- > > > Arnout Engelen > > > ASF Security Response > > > Apache Pekko PMC member, ASF Member > > > NixOS Committer > > > Independent Open Source consultant > > > > > > --------------------------------------------------------------------- > > > To unsubscribe, e-mail: dev-unsubscr...@pekko.apache.org > > > For additional commands, e-mail: dev-h...@pekko.apache.org > > > > > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: dev-unsubscr...@pekko.apache.org > > For additional commands, e-mail: dev-h...@pekko.apache.org > > > > >
