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
{{ message }}
Repository navigation
test(flagd): drop the per-scenario /stop in e2e teardown once the testbed includes flagd-testbed#396 #1891
ProviderSteps.tearDown (providers/flagd/src/test/java/.../e2e/steps/ProviderSteps.java) POSTs /stop to the launchpad after every scenario. That makes every /start a cold flagd start.
open-feature/flagd-testbed#396 makes /start reuse the running flagd and restore the baseline flag state in place. Isolation no longer needs the /stop, and keeping it means the suite never gets the speedup.
Proposal
After bumping the testbed to a release that includes open-feature/flagd-testbed#396, remove the /stop block from tearDown. Keep ContainerPool.release and the NoOp provider swap.
Evidence
Tested locally with the /stop removed, against an image built from flagd-testbed#396 and one built from testbed main. All 590 tests (in-process 224, RPC 214, file 152) passed in every run. Wall time (in-process / RPC / file):
with /stop
without /stop
flagd-testbed#396
102-107s / 73-79s / 48-65s
96s / 66s / 41s
testbed main
100s / 75s / 50s
120s / 79s / 49s
The machine was shared with other test runs, so treat the times as indicative. Without flagd-testbed#396, dropping /stop gains nothing, which is why this should wait for the bump.
Activity