[Bugfix][TENT] Replace stale buffers when refreshing segment info

[jacklin78911-collab]

Description

Reusing a SegmentInfo output object with getSegmentInfo() appends new buffers to the previous result. Repeated queries duplicate entries, and a refresh after unregistering a buffer still reports that old address.

Clear the output buffer list after acquiring the current segment descriptor, before filling either the memory or file result. This is a one-line runtime fix. Add a regression to the existing local-memory lifecycle test that queries twice into one object, unregisters the buffer, and queries again.

This fixes output-object reuse, not the remote descriptor cache. The existing TENT CI job already runs this test target. No new state, configuration or test target is needed.

Module

• Transfer Engine (mooncake-transfer-engine)

Type of Change

• Bug fix

How Has This Been Tested?

Built from main 2f13d400da4a3eb04d37263742866c1ce555d212, with USE_TENT=ON, USE_CUDA=OFF, and BUILD_UNIT_TESTS=ON.

cmake --build build --target tent_local_memory_lifecycle_test -j4
ctest --test-dir build --output-on-failure -R '^tent_local_memory_lifecycle_test$'

• Baseline regression fails: the second query reports two entries for one registered buffer, and querying after unregistration still returns entries.

• Fixed: both local-memory lifecycle tests pass, including the existing failed-free ownership regression.

• Pre-commit passes on the two changed files; git diff --check passes.

• Unit tests pass

• Manual baseline/fixed comparison done

• Upstream CI on bd3b2f645: all four TENT matrix jobs pass, including the local-memory lifecycle test suite in the CUDA-off job. The overall Build & Test run is red because the separate Store fileread_worker_pool_test fails (215/216 general CTests pass): process thread count grows beyond the test baseline, followed by the 1 s teardown-count timeout. This matches [CI] FilereadWorkerPoolTest.AcceptsTypedTrailingWhitespaceAndCaches is flaky on the 1s teardown window #4213; existing fixes [Bugfix][Store] Isolate fileread worker count assertions from runtime threads #4210 and [CI] Harden FilereadWorkerPool thread-count flake #4226 are still open. No Store test changes are included here. The author account cannot rerun upstream jobs (repository admin rights required).

Checklist

• Full diff and getSegmentInfo callers reviewed in a second pass
• C++ formatting checked through ./scripts/code_format.sh
• Scoped pre-commit passes
• Regression test added
• Human submitter has reviewed every changed line — draft pending that review

AI Assistance Disclosure

• AI tools were used

Codex assisted with diagnosis, implementation, regression validation and the second-pass review. This is a draft for human review; no human review is claimed.