| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
test: add checkpoint/loader_writer tests + report package tests - checkpoint: processCheckpoint branches, GC, InitFromCheckpoints, Maintain (46.8%) - report: usage_collector + usage_reporter via httptest/fake clientset (77.1%) Co-Authored-By: Claude <noreply@anthropic.com> | 12 天前 | |
test: fix CI-blocking numaaffinity test + relicense new tests (Huawei 2026) numaaffinity: TestInitNilClientSkipsWatcherAndTopologyDisabled failed in CI because it constructed the runner via NewNumaAffinityRunner(), which calls rest.InClusterConfig(). The CI runner is itself a pod with a service account, so InClusterConfig() succeeds there -> k8sClient becomes non-nil -> Init starts the ConfigMap watcher -> the assertion watcherStarted==false fails. Fix: construct &Runner{namespace: DefaultNamespace} directly so k8sClient is deterministically nil regardless of environment (Init never re-initializes the client). Verified by reproducing the in-cluster condition locally (service-account files + KUBERNETES_SERVICE_HOST/PORT) — the test now passes under the exact condition that failed in CI. No source change. Also update the license header on all 20 newly-added test files to the Huawei Technologies Co., Ltd. (2026) header, matching hack/boilerplate.go.txt, and gofmt-clean the previously-unformatted ones. Coverage unchanged at 83.0%; full suite passes with no failures. Co-Authored-By: Claude <noreply@anthropic.com> | 12 天前 | |
test: fix CI-blocking numaaffinity test + relicense new tests (Huawei 2026) numaaffinity: TestInitNilClientSkipsWatcherAndTopologyDisabled failed in CI because it constructed the runner via NewNumaAffinityRunner(), which calls rest.InClusterConfig(). The CI runner is itself a pod with a service account, so InClusterConfig() succeeds there -> k8sClient becomes non-nil -> Init starts the ConfigMap watcher -> the assertion watcherStarted==false fails. Fix: construct &Runner{namespace: DefaultNamespace} directly so k8sClient is deterministically nil regardless of environment (Init never re-initializes the client). Verified by reproducing the in-cluster condition locally (service-account files + KUBERNETES_SERVICE_HOST/PORT) — the test now passes under the exact condition that failed in CI. No source change. Also update the license header on all 20 newly-added test files to the Huawei Technologies Co., Ltd. (2026) header, matching hack/boilerplate.go.txt, and gofmt-clean the previously-unformatted ones. Coverage unchanged at 83.0%; full suite passes with no failures. Co-Authored-By: Claude <noreply@anthropic.com> | 12 天前 | |
test: fix CI-blocking numaaffinity test + relicense new tests (Huawei 2026) numaaffinity: TestInitNilClientSkipsWatcherAndTopologyDisabled failed in CI because it constructed the runner via NewNumaAffinityRunner(), which calls rest.InClusterConfig(). The CI runner is itself a pod with a service account, so InClusterConfig() succeeds there -> k8sClient becomes non-nil -> Init starts the ConfigMap watcher -> the assertion watcherStarted==false fails. Fix: construct &Runner{namespace: DefaultNamespace} directly so k8sClient is deterministically nil regardless of environment (Init never re-initializes the client). Verified by reproducing the in-cluster condition locally (service-account files + KUBERNETES_SERVICE_HOST/PORT) — the test now passes under the exact condition that failed in CI. No source change. Also update the license header on all 20 newly-added test files to the Huawei Technologies Co., Ltd. (2026) header, matching hack/boilerplate.go.txt, and gofmt-clean the previously-unformatted ones. Coverage unchanged at 83.0%; full suite passes with no failures. Co-Authored-By: Claude <noreply@anthropic.com> | 12 天前 |