The changelog in common.ai provider is technically right. (from commit history perspective)
``` Add branch_descriptions to LLMBranchOperator so the model reads what each branch means (#73639) ``` but it can be misleading as it's changed into branches later on (within 0.10.0rc1). Could we also change it before release? On 2026/09/28 10:40:34 Shahar Epstein wrote: > Thank you very much! > edge3 will be excluded from the current release (-1 binding). > > > Shahar > > On Mon, Sep 28, 2026 at 1:27 PM PoAn Yang <[email protected]> wrote: > > > Hi Shahar, > > > > Thanks for preparing this release! > > > > I found a regression in apache-airflow-providers-edge3 5.0.0rc1 that comes > > from my own PR #72048 [1]. When the scheduler restarts while Edge task > > instances from the previous scheduler are still queued or running, the new > > scheduler crashes on startup with DetachedInstanceError. It keeps crashing > > the same way on every restart for as long as those task instances stay > > queued or running. I reproduced it with the RC from PyPI on Airflow 3.3.2. > > The edge3 4.3.2 does not have the problem. I also tried the upgrade path: > > after a scheduler running edge3 4.3.2 queues some Edge tasks and stops, the > > first scheduler start after upgrading to 5.0.0rc1 crashes the same way. > > > > Could we exclude edge3? The fix is in [2]. Once that PR is merged, edge3 > > could go out in an rc2. > > > > +1 (non-binding) for amazon, apache.beam and google > > -1 (non-binding) for edge3 because of this regression > > > > [1] https://github.com/apache/airflow/pull/72048 > > [2] https://github.com/apache/airflow/pull/73832 > > > > Thanks, > > PoAn > > > > > > > On Sep 27, 2026, at 4:00 PM, Sameer Mesiah <[email protected]> wrote: > > > > > > +1 (non-binding) > > > > > > I can verify that my changes for the Amazon and Snowflake providers for > > > this release wave are working as expected. > > > > > > On Fri, 25 Sep 2026 at 09:50, Wei Lee <[email protected]> wrote: > > > > > >> +1 (binding), checked: > > >> > > >> - SVN > > >> - in Docker installation > > >> - Reproducible package builds > > >> - Licence > > >> - Signature > > >> - Checksums > > >> > > >> Thanks! > > >> > > >> Best regards, > > >> Wei Lee > > >> > > >> On 2026/09/25 08:22:48 Pankaj Koti via dev wrote: > > >>> +1 non-binding > > >>> > > >>> Tested that my PRs are included in the RC and working as expected, > > thanks > > >>> for cutting the release! > > >>> > > >>> On Fri, Sep 25, 2026 at 3:32 AM Jarek Potiuk <[email protected]> wrote: > > >>> > > >>>> +1 (binding) > > >>>> > > >>>> Checked: > > >>>> > > >>>> * SVN: all expected files present (check-release-files), and all 78 RC > > >>>> providers install at their RC versions (Dockerfile.pmc) > > >>>> * Signatures: all 157 good (Shahar's key from KEYS) > > >>>> * Checksums: all 157 SHA512 good > > >>>> * Licences: Apache RAT 0.18 - 0 unapproved, 0 unknown > > >>>> * Reproducible builds: the source tarball and all 156 packages are > > >>>> byte-identical when rebuilt from the providers/2026-09-22 tag with > > >>>> flit 4.0.2 > > >>>> > > >>>> One note on reproducibility, not a blocker: the wheels were built with > > >>>> flit 4.0.2, but dev/breeze/uv.lock at the tag locks flit 4.1.0. Since > > >>>> flit builds without isolation, rebuilding with the locked Breeze (as > > >>>> the docs say) gives 75 wheels that differ only in the "Generator:" > > >>>> line of their WHEEL file (and its hash in RECORD) - the packaged files > > >>>> are identical. I opened a PR explaining how to rebuild with the > > >>>> release manager's flit version: > > >>>> https://github.com/apache/airflow/pull/73699 > > >>>> > > >>>> I also tested all my changes in this wave with all 78 RC providers > > >>>> installed from the PyPI wheels in Breeze (Airflow 3.3.2): > > >>>> > > >>>> * #72162 (amazon, google, microsoft.azure) - remote log uploads > > outside > > >>>> base_log_folder are skipped, uploads inside still work > > >>>> * #72166 (google) - Drive q= values are escaped > > >>>> * #73413 (common.sql) - DataFusion error message shows the storage > > type > > >>>> * #72627 (edge3) - docs only, present in the release > > >>>> * #72174 (snowflake) and #72176 (teradata) - in the wave, though not > > >>>> listed in the testing issue - both work as expected > > >>>> > > >>>> Improvement to the verification process: > > >>>> > > >>>> I tested all my changes with an agent using Apache Magpie's > > >>>> release-management > > >>>> "verify-rc" skill and a new Airflow override that adds an optional > > >> "test > > >>>> your own changes" step. The agent found my entries in the testing > > issue > > >>>> (and the two changes missing from it), verified each change is in the > > >>>> RC tags, changelogs and wheels, downloaded the whole wave and wrote a > > >>>> Dag exercising every change - I only had to start Breeze and trigger > > >> it. > > >>>> It also ran the regular checks above and tracked down the flit > > >>>> difference. > > >>>> > > >>>> That was a very good experience - I highly recommend trying it: > > >>>> https://github.com/apache/airflow/pull/73697 > > >>>> > > >>>> J. > > >>>> > > >>>> --------------------------------------------------------------------- > > >>>> To unsubscribe, e-mail: [email protected] > > >>>> For additional commands, e-mail: [email protected] > > >>>> > > >>>> > > >>> > > >> > > >> --------------------------------------------------------------------- > > >> To unsubscribe, e-mail: [email protected] > > >> For additional commands, e-mail: [email protected] > > >> > > >> > > > > > --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
