davidzollo commented on PR #9898: URL: https://github.com/apache/seatunnel/pull/9898#issuecomment-5322956941
Both findings from the 2026-08-10 review are addressed on the new head: - **Issue 5 (blocker)**: `FileFormat.WORD` added to `BaseFileSource#buildCatalogTableForEmptyPath`'s self-describing list, so Cos/Obs/OssJindo Word sources started against an empty path now get the real 10-column schema instead of the simple-text fallback — with a comment pinning this list to its `BaseFileSourceConfig` sibling so the next self-describing format doesn't repeat the pattern a third time. Good catch on this one; it's exactly the kind of parallel-implementation gap that's invisible until the first gate is fixed. - **Issue 6**: all ten `WordReadStrategy` parsing/mapping helpers now carry method-level Javadoc documenting the non-obvious rules (paragraph-level run aggregation, comma-joined hyperlink URLs, the filtered `NONE` underline, the `000000` color default, the lossy `" | "`/newline table projection). On the CI note in section 3: the `unit-test (8, windows-latest)` failure on the previous head wasn't `WordReadStrategyTest` — I pulled the log, and the only error was `WorkerTagTest` timing out at 311s in `before()` (`Node failed to start!`), the roving Hazelcast/slow-Windows-runner flake hitting several unrelated PRs since the new windows-2025 runner image rolled out. `WordReadStrategyTest` passed on Windows in that same job. Also merged current dev — clean, no conflicts. -- 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]
