Gabriel39 commented on PR #68631: URL: https://github.com/apache/doris/pull/68631#issuecomment-5906263761
Fixed the Flight failures from P0 build 1060727 in f44c66c7aa4. Schema-only analysis created QueryState with isQuery=false, so result-sink binding inferred command-style names such as __literal_0 and __unhex_0. Real StmtExecutor execution sets isQuery=true and keeps SQL labels such as 1 and unhex(encoded). Preparation now sets query state before analysis, preserving the execution labels without weakening schema-drift checks. Validated before committing against a real isolated local FE + BE cluster, using this PR's CI BE binary and the actual regression framework: - Previous FE: all five affected Flight suites reproduced schema-drift failures; the new regression also failed its column-name assertion. - Fixed FE, same BE: test_flight_timestamp_range, test_flight_utf8_validation, test_sql_cache_over_arrow_flight, test_select, test_auth_remote_ip, and test_schema_only all passed (6 suites, 0 failures, 0 skipped). - All 60 focused FE tests and FE Checkstyle passed. The execution-schema unit tests now invoke the real FlightSqlConnectProcessor instead of calling the planner directly; three tests failed before the fix and passed afterward. The separately muted mtmv_with_sql_cache failure also occurs on unrelated PRs and is outside this Flight fix. The local test cluster has been stopped. Full CI has been requested separately. -- 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]
