abfs stuff has gone in to 3.2 -credit Thomas Marquardt and Da Zhou there. I'll do the same for S3A, including SDK updates And I'd like to pull back two changes to the FS APIs
HADOOP-15691 PathCapabilities. Lets apps probes for FS instances having specific features/semantics HADOOP-15229 Add FS/FC builder open() API. making the apis broadly available, even if any tuning in the object stores is left out, makes it possible to adopt. One thing to highlight which has cause problems on some backporting is the mockito update https://issues.apache.org/jira/browse/HADOOP-16275 without that mock tests can still compile -but they go on to fail.I think we will need to put that into 3.2 at the very least. Same for SLF4J backporting big patches to 3.1.x is harder because of the moves to slf4j spanning things: you may not have merge conflicts but things don't compile. On Tue, Oct 8, 2019 at 10:15 AM Wei-Chiu Chuang <weic...@apache.org> wrote: > I spent the whole last week cherry picking commits from trunk/branch-3.2 to > branch-3.1 (should've done this prior to 3.1.4 code freeze). There were > about 50-60 of them, many of them are conflict-free, and several of them > are critical bug fixes. > > If your commit stays in trunk, it'll be useless for the community until the > next minor release, and many months after people start using the new > release. > > Here are a few tips: > (1) update dependency to address a know security vulnerability, should be > cherry picked into all lower branches, especially when it updates the > maintenance release number. Example: update commons-compress from 1.18 to > 1.19. > > (2) blocker/critical bug fixes should be backported to all applicable > branches. > > (3) because of the removal of commons-logging and a few code refactors, > commits may apply cleanly but doesn't compile in branch-3.2, branch-3.1 and > lower branches. Please spend the time to verify a commit is good. > > Best > Weichiu >