Hello Jark and release team,
Thanks for your efforts!
+1 (non-binding) with a question below.
I verified the following on macOS arm64.
- Signatures
All six .asc files (src, bin, gateway amd64/arm64, helm chart, chart
.prov) verified with key E2C45417BED5C104154F341085BACB5AEFAE3202, which
is present in KEYS (dist/release KEYS identical to
downloads.apache.org/fluss/KEYS).
- Checksums
All .sha512 files OK.
- Source Release
The tag v1.0.0-rc1 resolves to b3b1be2ca6895f047d03be1d7b373f5983fe16a3;
the tarball matches that commit apart from the expected exclusions
(website/, .idea). There are no binaries or .git metadata in the
tarball.
- Build
Running `mvn clean package -DskipTests` with JDK 11 succeeds (RAT: 0
unapproved over 3235 files). The built fluss-dist matches the released
fluss-1.0.0-bin.tgz file for file.
- Helm Chart
Running `helm verify` (.prov) OK, `helm lint` clean, chart contents
identical to a `helm package` of the source tarball's helm/ directory,
`helm template` with `--set image.tag=1.0.0-rc1` renders
apache/fluss:1.0.0-rc1.
- Docker
Both images `apache/fluss:1.0.0-rc1` (index
sha256:465e263b8cf49eba439a4b67e14e4dfa620811c9e7b93b21d12c781570eaf531)
and `apache/fluss-quickstart-flink:1.20-1.0.0-rc1` carry linux/amd64 and
linux/arm64. The jar set in the fluss image equals the binary
distribution.
- Maven Staging
The fluss-client 1.0.0 jar and pom signatures/sha1 OK. Built our
smoke-test CLI against orgapachefluss-1013 and ran PK and log write/read
cycles (5000 rows each) against a local cluster started from
fluss-1.0.0-bin.tgz.
**Question on LICENSE/NOTICE file**:
The `apache/fluss:1.0.0-rc1` image has no LICENSE or NOTICE under
/opt/fluss (same as 0.9.1-incubating). The tarball contains them, so no
issue maybe. It is not blocking this release, but the Dockerfile (or the
image build step) should copy LICENSE-bin and NOTICE-bin so the image
matches the tarball.
Should I open an issue for 1.0.x?
**Other Notes**:
The fluss-client 1.0.0 on JDK 17 needs
`--add-opens=java.base/java.nio=ALL-UNNAMED` for the shaded Arrow
allocator (fails fast with a clear message otherwise; JDK 11 works as
is).
A standalone JDK 17 application using fluss-client 1.0.0 fails on first
write with "Failed to initialize MemoryUtil. You must start Java with
--add-opens=java.base/java.nio=ALL-UNNAMED" (shaded Arrow allocator).
Once adding the flag solves it, and JDK 11 works without it.
Since the docs recommend Java 17, we should mention the flag in the Java
client documentation page.
Best,
Muhammet
