Closed rbreslow closed 5 years ago
@hectcastro
Travis had a feature that allowed items in the build matrix to be marked such that their failures would not fail the overall build. It something like that exists for Circle, then I think using it for 2.12 makes sense.
This feature of Travis is called allow_failures
.
There are numerous threads on the CircleCI discussion boards and idea tracker that make it seem like this is not yet supported.
Also, I was able to get CircleCI build times (44s) to be faster than what I could achieve on my workstation (57s).
First, I deleted the target
directories, and then timed the tests.
maiden:maml rbreslow (feature/jrb/add-strta-for-ci)$ git clean -dxff
Removing js/
Removing jvm/target/
Removing project/project/project/
Removing project/project/target/
Removing project/target/
Removing spark/target/
Removing target/
maiden:maml rbreslow (feature/jrb/add-strta-for-ci)$ time ./sbt ++2.11.12 test
...
real 0m57.904s
user 4m25.443s
sys 0m10.524s
I also confirmed that, when the target
directories are gone, it's going to have to log something about resolving dependencies, but this is greatly truncated now that we're caching the right Coursier directory.
This looks good to me. The 2.12 bug appears to be resolved with PR #95
Overview
cibuild
andtest
STRTANotes
I had tried to add OpenJDK 11 to the build matrix, but was running into compatibility issues. I think we should split this out into a separate issue.
Scala 2.12.8 builds are also failing. I took a pass but I'm not immediately sure what's up. I'm inclined to leave Scala 2.12.8 in the build matrix, as it seems more important than supporting OpenJDK 11, but I'm not sure what the path forward should be.
Testing Instructions
See the CI workflow: https://circleci.com/workflow-run/ea13448f-59ae-4414-9b12-abe5252104a5