voonhous commented on PR #18816:
URL: https://github.com/apache/hudi/pull/18816#issuecomment-5464256165
Rebased onto master (the branch was 442 commits behind) and pushed. Green
locally: checkstyle clean, and `mvn test -Pfunctional-tests -pl hudi-cli`
passes with 8 skipped.
On top of your commit:
- Dropped three files that look accidental: `.sdkmanrc`,
`hudi-test-output.txt`, and one named `[pre-clean,`.
- Fixed 14 checkstyle `ImportOrder` violations. The `HoodieTableVersion`
import sat above the `org.apache.hudi.cli.*` group in five classes; that is
what `validate-source` was failing on.
- Three rebase conflicts resolved toward master, which had landed equivalent
fixes: `HoodieTestCommitMetadataGenerator` uses `metaClient.getTimelinePath()`
rather than a hardcoded `.hoodie/timeline/`, and `TestTableCommand` keeps the
static `CURR_VERSION` import.
- Narrowed the `RepairsCommand` corruption check and re-enabled two tests,
both below.
`RepairsCommand`: the gap is real, but I probed what actually gets thrown.
An empty `.clean.requested` produces a plain `java.io.IOException` with message
`unable to read commit metadata for instant [...]` -- not an `EOFException`,
and the message is not null. Of the five conditions only `contains("unable to
read")` ever fires, so I kept that one (matched against the fuller `"unable to
read commit metadata"`) plus the original Avro check and dropped the rest. The
`getMessage() == null` clause in particular would let any IOException delete an
instant file. It is still message matching, which is brittle -- a typed
exception from the timeline reader would be the real fix, but that is beyond
this PR.
`TestCompactionCommand`: removed the two `@Disabled`. Both `showarchived`
tests pass once the table is created at the current table version, so the
archived-timeline read was not what was failing. 5 passed, 0 skipped.
One question: this enables hudi-cli in Azure only. Codecov is fed by the
GitHub Actions jobs in `.github/workflows/bot.yml`, which still carries
`!hudi-cli` in `UT_MODULES` and `FT_MODULES`. Worth dropping it there too, so
the module reports real coverage instead of the current 15%?
--
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]
