You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We've observed non-deterministic behavior when running ginkgo:
for i in `seq 1 20`; do rm -rf /root/go/pkg; ginkgo -r -skipPackage=integration,brats; done
Produces a roughly 50% failure to compile with varying compilation errors.
for i in `seq 1 20`; do rm -rf /root/go/pkg; ginkgo -r -skipPackage=integration,brats -compilers=1; done
Succeeds 100% of the time.
Forcing module download before running tests, for example with go mod download, will also produce 100% success rate.
Our theory is that parallel compilation, and thus parallel module fetching, really confuses go, leading to races in module resolution/solving/whatever you call it.
I think this is surfacing a bug in go mod download. Seems like go should be able to handle multiple downloads in parallel. I'd suggest y'all open an issue there.
For now the -compilers=1 flag workaround will have to do :/
We've observed non-deterministic behavior when running ginkgo:
Produces a roughly 50% failure to compile with varying compilation errors.
Succeeds 100% of the time.
Forcing module download before running tests, for example with
go mod download
, will also produce 100% success rate.Our theory is that parallel compilation, and thus parallel module fetching, really confuses go, leading to races in module resolution/solving/whatever you call it.
@Zanadar
The text was updated successfully, but these errors were encountered: