+1 binding

On Wed, 16 Sep 2026 at 13:33, Amogh Desai <[email protected]> wrote:

> +1 binding.
>
> For both 3.3.2 airflow and 1.3.2 task sdk:
> - SVN checks
> - Reproducible package checks
> - License checks
> - Signature checks
> - Checksum checks
>
> Installed the release bits and validated my changes too, looks good.
>
>
> Thanks & Regards,
> Amogh Desai
>
>
> On Mon, Sep 14, 2026 at 2:50 AM Jarek Potiuk <[email protected]> wrote:
>
> > +0
> >
> > The artefacts themselves are in good shape — I ran the PMC mechanical
> > checklist and everything substantive passes:
> >
> > * reproducible build: all 7 artefacts (5 Airflow + 2 Task SDK) rebuild
> >   bit-for-bit identical from a clean clone of the 3.3.2rc1 tag
> > * signatures: 7/7 good, signing key present in the project KEYS file
> > * checksums: 7/7 sha512, no deprecated md5
> > * Apache RAT 0.18: 9591 files, 0 unapproved, 0 unknown
> > * NOTICE unchanged; LICENSE gains a correct OFL-1.1 section for the two
> >   bundled fonts, with both licence texts and both .woff2 files shipped
> > * versions consistent at 3.3.2 / 1.3.2 throughout
> >
> > My +0 is for source-release hygiene: two things ship in the source
> > tarball that should not, both of them already solved on main and simply
> > not carried over to v3-3-test.
> >
> > First, five dangling symlinks. All five are
> >
> >   .claude/skills/<name> -> ../../.agents/skills/<name>
> >
> > for aip-user-stories, airflow-translations, magpie-setup,
> > prepare-providers-documentation and upgrade-fab-provider.
> >
> > .gitattributes has ".agents export-ignore" but no ".claude
> export-ignore",
> > so the relay symlinks ship while the targets they point at are stripped.
> > Anyone unpacking the source release gets five broken links, in a .claude
> > directory that then contains nothing else at all.
> >
> > Fix is up as https://github.com/apache/airflow/pull/73107.
> >
> > Second, same category: the source release also contains
> > java-sdk/gradle/wrapper/gradle-wrapper.jar, a 43 KB compiled binary. ASF
> > policy does not permit compiled binaries in a source release and the
> > Gradle wrapper is not an exempted build tool (LEGAL-570), which is why
> > main stopped shipping it in #69444. Only half of that change was
> > backported to v3-3-test — the breeze side that restores gradlew is
> > there, the .gitattributes side is not — so the jar is still in this
> > tarball.
> >
> > Backport is up as https://github.com/apache/airflow/pull/73108,
> > which also starts shipping java-sdk/.editorconfig, stripped today even
> > though ktlint reads it at build time.
> >
> > To be clear, neither is a regression: both are in the released 3.3.1 as
> > well, so this RC introduced nothing, and neither breaks any part of the
> > build. That is why it is a +0 and not a -1 — I would not block the
> > release on either, but I would rather we not keep shipping them. Both
> > fixes are non-blocking and could ride along if an rc2 happens for other
> > reasons.
> >
> > Full verification detail is on the testing issue:
> > https://github.com/apache/airflow/issues/73089#issuecomment-5656115792
> >
> > Jarek
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: [email protected]
> > For additional commands, e-mail: [email protected]
> >
> >
>

Reply via email to