> FYI, devbranch (pre-commit) on ci-cassandra.a.o is hardcoded to build on > jdk8 and jdk11 > so it's not going to work with trunk based patches… > > I'll review any patches to fix this, but otherwise focus is > on CASSANDRA-18594 where a standalone jenkinsfile avoids this problem > altogether. >
To clarify this, devbranch was hardcoded to a matrix with jdk8 and jdk11, with a hack in place to skip jdk11 for <4 branches (at run time). Dropping jdk8 was not the cause of devbranch no longer working, it's just not built for it. These limitations of devbranch (and cassandra-builds) are being addressed as part of the repeatable ci work. Note, this does not affect post-commit ci-cassandra.a.o pipelines (which already is testing jdk11+17 on trunk). So… I would like to suggest we abandon CASSANDRA-18701 and we collectively focus on the intended replacement for this, which is the in-tree scripts and 18594. (It might possibly come as part of CASSANDRA-18665 which is an intermediate ticket). Stefan and Maxim, I think this impacts the two of you most over the next few weeks. In the meantime I suggest using the in-tree scripts as much as possible, using individual stage jobs under the patches tab in ci-cassandra.a.o (just ignoring the jdk8 axis failures), and asking others for final pre-commit ci runs on circleci. We're close.