superdiaodiao commented on PR #65698: URL: https://github.com/apache/doris/pull/65698#issuecomment-5599854590
Rebased and reduced this to test-only. Explanation of what changed and why: The FE fix this PR originally carried is obsolete. #67173 removed the cached `BaseTableStream.baseTable` field outright, so `getBaseTableNullable()` now resolves through `baseTableInfo.getTableNullable()` on every call and returns `null` once the base table is dropped; `TableStreamManager` then emits `N/A` for the four `BASE_TABLE_*` columns. My three-line guard patched a field that no longer exists, so I dropped it rather than rebase it forward. The branch is now current master plus one new regression test (+75/-0, one file). Two things I think are still worth acting on: 1. **#65389 is still open**, though its symptom appears fixed. #67173 closed #67095 and #67109; #65389 (`table_streams` still shows dropped base table name) was not linked, so the issue is outstanding even though the behaviour it reports is now correct. 2. **The path has no non-cloud regression coverage.** #67173 shipped with `test_olap_table_stream_schema_sync.groovy`, which covers schema sync rather than base-table removal. Under `table_stream_p0` nothing drops a base table and asserts the `table_streams` output; the only place `N/A` is asserted against `table_streams` is a `cloud_p0` e2e suite, and this test skips cloud mode. So this is the internal-mode guard against the cache being reintroduced. @TsukiokaKogane could you take a look? Pinging you specifically because you wrote #67173 and authored four of the last six commits to `BaseTableStream.java`, so this test is guarding behaviour you introduced — you're best placed to say whether the assertions match what you intended. One caveat I want to state rather than paper over: I have not been able to run the regression suite locally, so `run buildall` is what actually validates this. If it passes, the test pins current behaviour. If it fails, that is a more interesting result than a green run, because it would mean the dropped-base-table path regressed somewhere after #67173 — either way the outcome is worth having. Kicking off a run now. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
