Duplicate approvals

[JacobHayes]

After making a change to an approved path, every subsequent commit triggers a new approval. The logs say it is looking for existing reviews, but seems to not find the last one.

Logs from the second run:

Fetching user, pull request information, and existing reviews
Current user is github-actions[bot]
Commit SHA is AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
Pull request #123 has not been approved yet, creating approving review
Approved pull request #123

workflow config:

name: Auto approval
on:
  pull_request:
    branches:
      - main
    paths:
      - 'some-dir/**'
jobs:
  auto-approve:
    runs-on: ubuntu-latest
    permissions:
      pull-requests: write
    steps:
      - uses: hmarr/auto-approve-action@v3
        with:
          review-message: "Automatic approval of ..."

---

Perhaps there's a missing permission that is preventing it from fetching existing reviews?

[hmarr]

I believe the default behaviour is to re-approve when new commits are pushed to the branch (commit_id == prHead):

auto-approve-action/src/approve.ts

Lines 39 to 45 in aa8a692

	const prHead = pr.head.sha;
	core.info(`Commit SHA is ${prHead}`);
	const alreadyReviewed = reviews.some(
	({ user, commit_id, state }) =>
	user?.login === login && commit_id == prHead && state === "APPROVED"
	);

I assume that's so that re-reviews occur when they are required by branch protections:

However, the state === "APPROVED" check should cover that case, as when that box is checked, existing reviews transition to DISMISSED when new commits are pushed.

I'll try removing the check in a PR.

[hmarr]

Released in v3.2.0

[pfuhrmann]

I do not believe this works correctly now.

I have DISMISSED reviews, and the action is not re-approving PRs (it states that the PR was already approved by the user, which is true, but the review is dismissed and it seems to not detect that state).

Previously it worked correctly (albeit yes, it was duplicating sometimes, but I did not mind that).

[hmarr]

Thanks @pfuhrmann. It looks like it's possible to have old APPROVED reviews with newer DISMISSED reviews, and only the latest review for a given user counts towards a PR's review state, which is why this is happening. I've opened #216 to fix that.

[hmarr]

@pfuhrmann if you could confirm that you're seeing the same thing (old APPROVED reviews with newer DISMISSED reviews for a given user), that'd be helpful.

You can see all the reviews by running gh api repos/{LOGIN}/{REPO}/pulls/{PR-NUMBER}/reviews if you have the gh cli installed.

[pfuhrmann]

Hi @hmarr, thanks for coming back to me so quickly!

This is what I see on the PR (redacted):

[
  {
    "id": 1340392942,
    "node_id": "PRR_kwDOH65R-s5P5MXu",
    "user": {
      "login": "github-actions[bot]",
    },
    "body": "Auto approved (low risk update)! :+1: :rocket:",
    "state": "APPROVED",
    "submitted_at": "2023-03-14T22:41:54Z",
    "commit_id": "2e0af399a670ff3971253f69539274b1dedf13a1"
  },
  {
    "id": 1340392945,
    "node_id": "PRR_kwDOH65R-s5P5MXx",
    "user": {
      "login": "github-actions[bot]",
    },
    "body": "Auto approved (low risk update)! :+1: :rocket:",
    "state": "DISMISSED",
    "submitted_at": "2023-03-14T22:41:55Z",
    "commit_id": "2e0af399a670ff3971253f69539274b1dedf13a1"
  }
]

The full output from action is:

Fetching user, pull request information, and existing reviews
Current user is github-actions[bot]
Commit SHA is 2968fc1548df680b12c727068ad3f9075c56294f
Current user already approved pull request #77, nothing to do

I assume this is the state you wrote about previously, i.e. old APPROVED reviews with newer DISMISSED reviews. Seem's DISMISSED PR submitted_at is 1s later than APPROVED. Probably an edge case, and can happen if there are multiple actions running in parallel?

[hmarr]

Great, yeah that does look like the case I described. I've merged the PR and released v3.2.1 — hopefully that fixes the issue for you!

[pfuhrmann]

@hmarr I can confirm, all our automated PRs got merged so this seems resolved. Thanks!