potiuk commented on issue #73089:
URL: https://github.com/apache/airflow/issues/73089#issuecomment-5656115792
### Verification of 3.3.2rc1 / Task SDK 1.3.2rc1
I ran the PMC mechanical checks from `dev/README_RELEASE_AIRFLOW.md` against
the staged artefacts (svn r87505). Result: **pass, with two non-blocking
warnings that are both pre-existing in the released 3.3.1**.
| Check | Result |
|---|---|
| Artefact inventory | PASS — 15 Airflow files + 6 Task SDK, nothing missing
or unexpected |
| GPG signatures | PASS — 7/7 good, signing key present in the project KEYS
file |
| Checksums | PASS — 7/7 sha512, no deprecated md5 |
| Licence headers (Apache RAT 0.18) | PASS — 9591 files, 0 unapproved, 0
unknown |
| NOTICE / LICENSE | PASS — NOTICE unchanged vs 3.3.1; LICENSE +20 lines,
intentional (#73067) |
| Version consistency | PASS — `3.3.2` / `1.3.2` throughout |
| **Reproducible build** | **PASS — all 7 artefacts bit-for-bit identical** |
Reproducibility was checked from a clean `git clone --branch 3.3.2rc1` using
the default Docker build path. All seven artefacts compare byte-identical to
the staged ones.
The LICENSE addition is correct and complete: the OFL-1.1 section added by
#73067 covers JetBrains Mono and Plus Jakarta Sans, both `LICENSE-*.txt` files
ship in the source release, and both `.woff2` files they cover are present.
(`codicon.ttf` in the core wheel is a Monaco Editor asset, already covered by
the existing MIT entry — not a gap.)
#### Two warnings — neither blocks the vote
**1. Five dangling symlinks in the source release.**
`.claude/skills/{aip-user-stories,airflow-translations,magpie-setup,prepare-providers-documentation,upgrade-fab-provider}`
all point at `../../.agents/skills/...`. `.gitattributes` has `.agents
export-ignore` but no `.claude export-ignore`, so the relay symlinks ship while
their targets are stripped. Present since #67929 and identical in the released
3.3.1. One-line fix for a follow-up PR on main, not for this RC.
**2. `java-sdk/gradle/wrapper/gradle-wrapper.jar` (43 KB) in the source
release.** RAT classifies it as an archive and does not complain. Identical at
the `3.3.1` tag. Worth a policy decision eventually — ASF prefers source
releases free of compiled binaries — but not grounds to -1 here.
#### The 23 items assigned to me
All 23 are in the `3.3.2rc1` tag, and all 23 are new in 3.3.2.
Nine of them — #70597, #71428, #71434, #71451, #71512, #71640, #71785,
#71828, #71838 — do **not** appear under the merge-commit SHA the API reports,
because `v3-3-test` was rebased/force-pushed after they merged. I re-confirmed
each by matching the PR reference in the tag history and comparing `git
patch-id --stable` between the original and rewritten commits: all nine carry
byte-identical patches. The other fourteen (#72660, #72712, #72723, #72728,
#72806, #72874, #72933, #73041, #73042, #73043, #73046, #73051, #73067, #73068)
are direct ancestors of the tag.
Worth noting that #73068 (RAT excludes, move to Apache RAT 0.18) is itself
in this RC, and the RAT run above used the 0.18 image it pins — so that fix is
verified against the artefact it was meant to fix.
These checks are mechanical and say nothing about runtime behaviour; an
install-and-exercise pass is still needed for a binding vote.
---
Drafted-by: Claude Opus 5; reviewed by @potiuk before posting
--
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]
