Yicong-Huang commented on code in PR #8096: URL: https://github.com/apache/texera/pull/8096#discussion_r3921646516
########## .github/workflows/backport-approval-check.yml: ########## @@ -0,0 +1,316 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +# Blocks the merge while a `release/*` label on the PR has not been approved by +# that branch's release manager (.github/release-branches.yml). +# +# The point is that a merged PR's labels should be a truthful record. A label +# only nominates a branch, so before this check a manager declined a backport by +# staying silent — and the label stayed on, which read afterwards as "this +# shipped in 1.2" when it had not. Here declining is an action: the manager +# either approves, or removes their label. Since the merge waits for every +# remaining label to be approved, the labels on a merged PR are exactly the +# branches Direct Backport Push then sends the fix to. +# +# The author's job is to ask the manager for that call, not to make it for them. +# Editing labels needs triage access, so outside contributors and Renovate or +# Dependabot PRs cannot remove a label themselves — the failure report says so +# and names who to ask. +# +# `Backport Approvals` is listed in .asf.yaml's required_status_checks, so this +# job must produce a result on EVERY pull request, or a PR that never runs it +# waits forever. That is why it is not conditioned on the title, the labels, or +# the base branch: it always runs and reports success when there is nothing to +# approve. Merge groups are re-evaluated rather than waved through, because a +# queued PR's own checks are no longer consulted once it is queued; a group +# whose PRs cannot be resolved fails closed rather than passing unchecked. Do +# not rename the job — its display name is the required context. + +name: Backport Approval Check + +on: + pull_request: + types: + - opened + - reopened + - synchronize + - labeled + - unlabeled + # Retargeting arrives as `edited`, and the verdict depends on the base + # branch: a backport PR into release/* has nothing to approve, the same + # PR moved onto main does. + - edited + pull_request_review: + types: + - submitted + - dismissed + merge_group: + +# Read-only: this job never writes to the PR, and `pull_request_review` runs +# with the base repository's token even for fork PRs. +permissions: + contents: read + pull-requests: read + +concurrency: + group: backport-approvals-${{ github.event.pull_request.number || github.ref }} + cancel-in-progress: true + +jobs: + backport-approvals: + # Do not rename — this display name is the required status check context + # referenced in .asf.yaml. + name: Backport Approvals + runs-on: ubuntu-latest + steps: + # The PR's own head (or the queued group's), so the config is the one that + # will govern this change once it merges. Checking out the base would not + # be a security boundary anyway: on `pull_request` the workflow definition + # itself comes from the PR. What bounds this is the event — a fork gets a + # read-only token and no secrets, and nothing here writes to the PR. + - name: Checkout + uses: actions/checkout@v7 + with: + ref: ${{ github.event.pull_request.head.sha || github.event.merge_group.head_sha }} Review Comment: This pins the PR head, and `:93` runs `release_branches.py` against that tree — so the gate reads its authorization config from a revision that can be arbitrarily old. `backport-auto-label.yml:79`, which creates the labels this gate judges, pins `base.sha` instead, under a comment saying the config must come from the base repo. `direct-backport-push.yml` reads main. Two of the three consumers agree; this one does not. Neither file exists on a head predating `31f1e7bf8` (#6941), nor on one branched from `release/v1.1` or `v1.2`. `run:` runs under `bash -eo pipefail`, so the missing file exits the step 2 and the required check goes red with a file-not-found instead of your report. **37 of the 222 open PRs into `main` are in that state** (list in the summary block), and none can clear it by approving or unlabeling. ```suggestion ref: ${{ github.event.pull_request.base.sha || github.event.merge_group.head_sha }} ``` The tradeoff: a PR that itself edits `release-branches.yml` would then be judged by the pre-edit config. That looks negligible here, since such PRs are `ci:`-typed and carry no `release/*` labels. `refs/pull/{n}/merge` keeps both but does not exist for a conflicting PR. -- 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]
