github actions: preserve open CLK PRs during lt-rebase-merge by bmastbergen · Pull Request #1341 · ctrliq/kernel-src-tree

bmastbergen · 2026-06-12T20:39:30Z

When lt_rebase_merge.sh runs, it deletes and recreates the ciq-6.X.y branch, which causes GitHub to auto-close any open PRs targeting it.

Add park/restore steps around the script invocation in the workflow: before the script runs, open PRs are retargeted to a temporary parking branch; after it succeeds, they are moved back to the new ciq-6.X.y.

On failure, PRs remain on the parking branch (still open) and a comment explains the situation. The parking branch is uniquely named with the run ID to avoid collisions with stale branches from prior failures.

Both transitions leave a comment on affected PRs so authors know what happened and that they need to rebase.

Copilot

Pull request overview

This workflow update aims to prevent GitHub from auto-closing open PRs when lt_rebase_merge.sh deletes and recreates the ciq-X.Y.y base branch by temporarily retargeting affected PRs to a per-run “parking” branch and restoring them afterward.

Changes:

Add a “park PRs” step that creates a temporary branch and retargets open PRs before running lt_rebase_merge.sh.
Add a “restore PRs” step that retargets parked PRs back to the recreated branch on success (or leaves them parked on failure) and posts explanatory comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

+          cd kernel-src-tree
+          git push origin "origin/${TARGET_BRANCH}:refs/heads/${TEMP_BRANCH}"
+          echo "TEMP_BRANCH=$TEMP_BRANCH" >> $GITHUB_ENV


+        run: |
+          set -e
+
+          OPEN_PRS=$(gh pr list --repo "$REPOSITORY" --base "$TARGET_BRANCH" --state open --json number --jq '[.[].number | tostring] | join(" ")')


+            exit 0
+          fi
+
+          if ! PARKED_PRS=$(gh pr list --repo "$REPOSITORY" --base "$TEMP_BRANCH" --state open --json number --jq '[.[].number | tostring] | join(" ")'); then


+        run: |
+          if [ -z "$TEMP_BRANCH" ]; then
+            echo "No temp branch, nothing to restore"
+            exit 0
+          fi


+            git fetch origin --tags --quiet 2>/dev/null || true
+            NEW_VERSION=$(git describe --tags --abbrev=0 --match='v*' "origin/${TARGET_BRANCH}" 2>/dev/null || echo "the latest upstream release")
+


+      - name: Park open PRs on temporary branch
+        env:
+          GH_TOKEN: ${{ steps.generate-token.outputs.token }}
+          REPOSITORY: ${{ github.repository }}
+          RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
+        run: |


PlaidCat · 2026-06-12T21:13:16Z

+                continue
+              fi
+              gh pr comment "$pr" --repo "$REPOSITORY" --body-file - <<EOF || true
+:white_check_mark: \`${TARGET_BRANCH}\` has been rebased to \`${NEW_VERSION}\`. This PR has been retargeted to the updated branch. You will need to rebase your PR branch onto the new \`${TARGET_BRANCH}\`.


Awesome.

Would love to see an auto attempt at re-basing for the PR, however this is beyond this PR

PlaidCat

there are some minor comments but I think this is a great MVP to solve something very annoying.

shreeya-patel98 · 2026-06-15T11:30:02Z

Some of the comments from claude review, feel free to discard the ones which don't make sense :-

The park step lists --base "$TARGET_BRANCH" (ciq-X.Y.y), but the PRs going through this workflow target -next (per the validation up top). Which branch does the script actually recreate, and which one do the open PRs sit on? If they're on -next, the park query finds nothing. Worth nailing down before this runs for real.
Partial park failure: TEMP_BRANCH gets written to $GITHUB_ENV before the retarget loop, so if a gh pr edit dies mid-loop under set -e, some PRs are parked and some aren't — and the un-parked ones can still get auto-closed. Maybe tolerate per-PR failures like the restore loop does.
Stale parking branches never get cleaned up — if a prior run failed and left PRs on _pr_parking_..._run<old>, the next run won't see them (it only queries $TARGET_BRANCH), so they're stuck there forever.

2 and 3 are edge cases, not blockers.

When lt_rebase_merge.sh runs, it deletes and recreates the ciq-6.X.y branch, which causes GitHub to auto-close any open PRs targeting it. Add park/restore steps around the script invocation in the workflow: before the script runs, open PRs are retargeted to a temporary parking branch; after it succeeds, they are moved back to the new ciq-6.X.y. On failure, PRs remain on the parking branch (still open) and a comment explains the situation. The parking branch is uniquely named with the run ID to avoid collisions with stale branches from prior failures. Both transitions leave a comment on affected PRs so authors know what happened and that they need to rebase.

bmastbergen · 2026-06-15T14:40:03Z

Some of the comments from claude review, feel free to discard the ones which don't make sense :-

The park step lists --base "$TARGET_BRANCH" (ciq-X.Y.y), but the PRs going through this workflow target -next (per the validation up top). Which branch does the script actually recreate, and which one do the open PRs sit on? If they're on -next, the park query finds nothing. Worth nailing down before this runs for real.

This workflow is running for PRs against ciq-6.x.y-next (ie. rebase PRs), but the PRs we need to park/restore are PRs targeting ciq-6.x.y. Sort of confusing, but correct.

Partial park failure: TEMP_BRANCH gets written to $GITHUB_ENV before the retarget loop, so if a gh pr edit dies mid-loop under set -e, some PRs are parked and some aren't — and the un-parked ones can still get auto-closed. Maybe tolerate per-PR failures like the restore loop does.

I think this is best effort. If for some reason we only manage to park/restore some PRs, thats still better than today.

Stale parking branches never get cleaned up — if a prior run failed and left PRs on _pr_parking_..._run<old>, the next run won't see them (it only queries $TARGET_BRANCH), so they're stuck there forever.

"Stuck there forever" as far as no workflow is going to do anything about them automatically. But there will still be open PRs that we see out there so we'll know they need to be dealt with. I figured we'd run like this for a while to actually see how things work before we start trying to do automated cleanup of any kind.

2 and 3 are edge cases, not blockers.

PlaidCat

Copilot AI review requested due to automatic review settings June 12, 2026 20:39

bmastbergen requested review from PlaidCat, kerneltoast and shreeya-patel98 as code owners June 12, 2026 20:39

bmastbergen requested a review from roxanan1996 June 12, 2026 20:39

Copilot started reviewing on behalf of bmastbergen June 12, 2026 20:39 View session

Copilot AI reviewed Jun 12, 2026

View reviewed changes

PlaidCat reviewed Jun 12, 2026

View reviewed changes

Comment thread .github/workflows/lt-rebase-merge.yml Outdated

PlaidCat reviewed Jun 12, 2026

View reviewed changes

Comment thread .github/workflows/lt-rebase-merge.yml Outdated

PlaidCat reviewed Jun 12, 2026

View reviewed changes

PlaidCat previously approved these changes Jun 12, 2026

View reviewed changes

bmastbergen dismissed PlaidCat’s stale review via 64d8f51 June 15, 2026 14:28

bmastbergen force-pushed the {bmastbergen}_main branch from 7587146 to 64d8f51 Compare June 15, 2026 14:28

PlaidCat approved these changes Jun 15, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

github actions: preserve open CLK PRs during lt-rebase-merge#1341

github actions: preserve open CLK PRs during lt-rebase-merge#1341
bmastbergen wants to merge 1 commit into
mainfrom
{bmastbergen}_main

bmastbergen commented Jun 12, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

PlaidCat Jun 12, 2026 •

edited

Loading

Uh oh!

PlaidCat left a comment

Uh oh!

shreeya-patel98 commented Jun 15, 2026 •

edited

Loading

Uh oh!

bmastbergen commented Jun 15, 2026 •

edited

Loading

Uh oh!

PlaidCat left a comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

4 participants

		git fetch origin --tags --quiet 2>/dev/null \|\| true
		NEW_VERSION=$(git describe --tags --abbrev=0 --match='v*' "origin/${TARGET_BRANCH}" 2>/dev/null \|\| echo "the latest upstream release")

Conversation

bmastbergen commented Jun 12, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

PlaidCat Jun 12, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

PlaidCat left a comment

Choose a reason for hiding this comment

Uh oh!

shreeya-patel98 commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bmastbergen commented Jun 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

PlaidCat left a comment

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

4 participants

PlaidCat Jun 12, 2026 •

edited

Loading

shreeya-patel98 commented Jun 15, 2026 •

edited

Loading

bmastbergen commented Jun 15, 2026 •

edited

Loading