Fix flaky TestEventServer_GetJobSetEvents context budget

[dejanzele]

TestEventServer_GetJobSetEvents_ErrorIfMissing (4 subtests) and TestEventServer_GetJobSetEvents_Permissions (3 subtests) share a single 5s armadacontext.WithTimeout across all their subtests. Each subtest spins up a fresh lookout DB and runs every migration via withEventServer and WithLookoutDb, so the per-subtest cost is not trivial.

On a slow CI runner the cumulative cost blew the 5s budget. We hit this on PR #4920 where 4 subtests of _ErrorIfMissing took 5.74s combined and the last one died with context deadline exceeded inside PostgresQueueRepository.upsertQueue.

[greptile-apps]

Greptile Summary

This PR fixes flaky tests in TestEventServer_GetJobSetEvents_ErrorIfMissing and TestEventServer_GetJobSetEvents_Permissions by moving the armadacontext.WithTimeout call inside each t.Run block, giving every subtest its own independent 5-second deadline instead of sharing a single one across all subtests.

• The shared outer context is removed; each t.Run closure now creates and defers its own ctx/cancel, so slow CI runners can no longer exhaust the budget before later subtests start.
• Both test functions are simultaneously refactored to table-driven style, extracting a reusable jobRunAssignedEvent() helper and eliminating ~100 lines of repetitive setup code.

Confidence Score: 5/5

This change only touches test code; no production logic is affected and the fix is structurally sound.

Each subtest now creates its own independent 5-second context inside its t.Run closure, directly addressing the cumulative-budget exhaustion described in the PR. The table-driven refactor reduces duplication without introducing new test logic. The defer cancel() is correctly scoped to the closure, loop-variable capture is safe (subtests are sequential), and no logic regressions were found across the four _ErrorIfMissing and three _Permissions cases.

No files require special attention.

Important Files Changed

Filename	Overview
internal/server/event/event_test.go	Per-subtest context isolation correctly implemented; table-driven refactor is clean with no logic regressions found.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A["TestEventServer_GetJobSetEvents_ErrorIfMissing"] --> B["for name, tc := range tests"]
    B --> C["t.Run(name, ...)"]
    C --> D["ctx, cancel := WithTimeout(5s)  ← NEW per-subtest"]
    D --> E["defer cancel()"]
    E --> F["withEventServer(ctx, t, ...)"]
    F --> G{"tc.publishEvent?"}
    G -- yes --> H["reportPulsarEvent(ctx, ...)"]
    G -- no --> I["GetJobSetEvents(...)"]
    H --> I
    I --> J{"tc.expectErrorCode == OK?"}
    J -- yes --> K["require.NoError"]
    J -- no --> L["assert gRPC error code"]
    K --> M["assert len(stream.sendMessages)"]
    L --> M

Loading

_{Reviews (8): Last reviewed commit: "Merge branch 'master' into fix-event-tes..." | Re-trigger Greptile}

[dejanzele]

@greptileai