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
Claude's follow-up review on PR #644 noted that
TestDynamoHandler_ForwarderNotInvokedForNonNotLeaderError only
exercised the create side, leaving tryForwardDelete's matching
guard untested. The DELETE forward error paths
(ErrLeaderUnavailable, generic transport error) also had no
direct coverage. Close the gap so a typo-class change to
tryForwardDelete or writeForwardFailure on the delete side fails
a test.
Three additions:
- Extend TestDynamoHandler_ForwarderNotInvokedForNonNotLeaderError
to a two-axis sweep: 3 create cases (already_exists, validation,
generic) × 3 delete cases (not_found, forbidden, generic) — 6
sub-tests pinning that the forward path runs ONLY on
ErrTablesNotLeader.
- TestDynamoHandler_ForwarderLeaderUnavailableReturns503_Delete:
503 + Retry-After: 1 + leader_unavailable body when the
forwarder reports election in flight.
- TestDynamoHandler_ForwarderTransportErrorReturns503_Delete:
503 + Retry-After: 1 + no internal detail leaking to the SPA
when the forwarder gRPC transport fails.
All three reuse the existing stubLeaderForwarder + notLeaderSource
fixtures — no new test scaffolding needed.
0 commit comments