voonhous commented on PR #19161: URL: https://github.com/apache/hudi/pull/19161#issuecomment-5291523827
Pushed ab6a61c addressing the remaining findings so this can move forward: - **Restored `TestExportInstantsProcedure` to master.** The branch was still reverting the desc-ordering / action-filter / invalid-folder tests from #19172 (the prod file was restored earlier, but the test file was not), so merging would have dropped the regression coverage for the `desc` bug. The PR no longer touches export_instants at all. - **Fixed the incremental `copy_to_temp_view` test.** The comment had the boundary semantics inverted: both bounds are inclusive on *completion* time, and passing the last commit's *requested* time as `end_instance_time` excludes that commit. The surviving row is the first commit's, and the test now asserts `id = 1` instead of a bare count. - **Made the filter-evaluator tests pin two real bugs instead of skirting them.** Long columns compared against int literals are narrowed to Int, so values past Int range silently never match -- pinned with a `3000000000L` row, tracked in #19632. Array predicates drop every row because values are never wrapped as `ArrayData` -- the tautological `length <=` assertion is now explicit `Seq.empty` pins, tracked in #19633. - **Documented the `show_logfile_records` workarounds.** The merge test's compaction scaffolding and distinct-key updates are required because the merged scan crashes on an uncompacted MOR table and on keys updated in more than one log block; tracked in #19634 together with the `MatchError` on non-data blocks. - **Filed #19635** for two `sync_validate` bugs adjacent to the new coverage (catch-up counting hardcodes the commit action, breaking MOR/clustered targets; an unguarded `conn.close()` masks JDBC failures with an NPE). No test change needed there. Everything stays test-only; the prod fixes live in the four issues above. -- 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]
