HyukjinKwon opened a new pull request, #59807:
URL: https://github.com/apache/spark/pull/59807

   ### What changes were proposed in this pull request?
   
   This PR makes the release workflow (`.github/workflows/release.yml`) always 
run master's `dev/free_disk_space` instead of the copy in the branch being 
released, as SPARK-55115 did for `Dockerfile.base`. The previous step already 
fetches `origin/master`, so the step now checks out master's script and runs it.
   
   ### Why are the changes needed?
   
   The `ubuntu-latest` label started migrating to Ubuntu 26.04 on October 19 
(https://github.com/actions/runner-images/issues/14748), and `apt-get` there 
can no longer locate the `powershell` package. Master's script tolerates this 
since SPARK-58934, which landed only in `master` and `branch-4.x`. The copies 
in `branch-4.1`, `branch-4.2` and `branch-4.3` still run the following under 
`set -e`, and the release workflow has no `continue-on-error` on that step:
   
   ```
   sudo apt-get remove --purge -y azure-cli google-cloud-sdk mono-devel 
powershell libgl1-mesa-dri
   ```
   
   So releasing these branches fails before the release starts, e.g., a 
`branch-4.3` run in 
https://github.com/HeartSaVioR/spark/actions/runs/40288893806/job/121589366145:
   
   ```
   + sudo apt-get remove --purge -y azure-cli google-cloud-sdk mono-devel 
powershell libgl1-mesa-dri
   E: Unable to locate package powershell
   Package 'google-cloud-sdk' is not installed, so not removed
   ##[error]Process completed with exit code 100.
   ```
   
   The script only prepares the GitHub runner, which is the same 
`ubuntu-latest` for every branch released, so it should come from master, which 
is maintained against the current image. Backporting SPARK-58934 would also fix 
this one case, but the release branches' copies would break again on the next 
image change. Master's script already works on Ubuntu 26.04: the scheduled dry 
run in `apache/spark` hits the same `Unable to locate package powershell`, 
tolerates it, and passes, e.g., 
https://github.com/apache/spark/actions/runs/40171636156.
   
   This does not affect release artifacts: the release builds from a fresh 
clone inside Docker, and only `dev/free_disk_space` is taken from master.
   
   ### Does this PR introduce _any_ user-facing change?
   
   No, dev-only.
   
   ### How was this patch tested?
   
   Dispatched this workflow from this branch in my fork as a `branch-4.3` dry 
run: https://github.com/HyukjinKwon/spark/actions/runs/40416194036. The "Free 
up disk space" step, which fails in the run above, passed. The rest of the dry 
run was still running when this PR was opened.
   
   Also ran the workflow's git steps on a fresh shallow `branch-4.3` checkout 
to confirm the checked-out `dev/free_disk_space` is executable and identical to 
master's.
   
   ### Was this patch authored or co-authored using generative AI tooling?
   
   Generated-by: Claude Code (Opus 5.5)
   
   This pull request and its description were written by Isaac.
   


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to