Repository navigation
Conversation
dgryski
left a comment
There was a problem hiding this comment.
In principle this is fine but our CI servers are already pretty overloaded. I'd like to hear from @deadprogram first before approving this.
|
I thought it was just macOS that was overloaded? I think adding more parallel Linux jobs does generally work? |
macOS is the most overloaded, but nearly all runners get queued these days. We are getting too much done! 😛 |
deadprogram
left a comment
There was a problem hiding this comment.
Thanks for working on this @rdon-key! I ran the tests on my local machine and looked it over. The following is edited from an automated review:
-
Since this branch was made, dev added a
coreassign.gotest with a skip that still usesspec.Scheduler. The branch merges cleanly, but after the mergeTestBuild/EmulatedRISCV/tasks/coreassign.gofails because the skip seescoresfrom the target spec:-second AssignCore panics: true +second AssignCore panics: falsePlease rebase on dev and change that check to use the new
schedulervariable too:if name == "coreassign.go" && scheduler != "cores" {
With that change, all of
TestBuild/EmulatedRISCVpasses for me with both schedulers.
What do you think of this?
ce924e8 to
22d7300
Compare
|
Thanks for catching this, Ron. You're right, I missed the new coreassign.go test added to dev. Thanks for testing this on your side as well! |
Fixes #5896.
Extend
TestBuild/EmulatedRISCVto run with both thecoresandtasksschedulers.Previously, only the default
coresscheduler was tested.Changes
EmulatedRISCVtests separately withcoresandtasks.goroutines.gotest exclusion to use the effective scheduler, respectingoptions.Schedulerwhen explicitly specified.Testing
Tested on Windows 11 with QEMU 11.1.0 and LLVM 22.1.4.
All tests passed with both schedulers.
cores: PASStasks: PASS (includinggoroutines.go)