Self hosted gitlab can not process follow up push

Where does the bug appear (feature/product)?

BugBot

Describe the Bug

We setup the self hosted gitlab connection with cursor, then created an MR to trigger the bugbot review, the bugbot created several comments, everything works fine. Then we try to resolve the comments, pushes more fixes, from here, bugbot seems to be stuck, no more actions shown in the MR page.

I tried to trigger the bot manual by the command: “cursor review verbose=true”, I only got one message as below

Bugbot request id: serverGenReqId_56aab9f5-cb89-421d-9a24-1072826c758d

Then there is no any other messages, seems like a server stuck, or some exceptions happened, the process got interrupt.

Also noted, we didn’t set the automatically run only once feature, the switch is off for that, both for gitlab config and user config, this switch is off. And the manual triggered run is not working, prove that this is not a regular config issue. And the first push triggered bugbot and the comments are successfully posted prove that the connection between self hosted gitlab and cursor is all good, so I think this should be a server bug.

Steps to Reproduce

  1. create a MR in self hosted gitlab
  2. wait for bugbot to review
  3. resolve the comment and push new commits
  4. there is no more actions from bugbot

Expected Behavior

the bugbot continue working.

Operating System

MacOS

Current Cursor Version (Menu → About Cursor → Copy)

this is a bugbot issue, no cursor version

For AI issues: add Request ID with privacy disabled

Bugbot request id: serverGenReqId_56aab9f5-cb89-421d-9a24-1072826c758d

Does this stop you from using Cursor

Yes - Cursor is unusable

Hey, thanks for the detailed report. Looks like an issue with handling webhook events from self-hosted GitLab after the first review.

Thanks for the Request ID - I’ll pass it to the engineers.

Can you check the following:

  1. In your GitLab project settings - Settings > Webhooks > check the Cursor webhook. Are there errors in Recent Deliveries?

  2. In webhook settings, check which events are enabled - Push events and Merge request events should be on.

  3. After a new commit - does a new webhook delivery appear in GitLab logs?

This will help us figure out where the connection breaks. I’ll pass the Request ID to the team - they’ll check logs on our side.

1 Like

Thanks for your reply, seems like we do not have a page showing the deliveries results, but I believe the delivery is success since when we run cursor review verbose=true, we do see a response from cursor bot with a resuqest id, so that id might be a good start for cursor team to investigate?

Then here is the screenshot of the webhook config for the cursor bugbot

1 Like

Thanks for the info.

If there’s no “Recent Deliveries” page in GitLab, we can check via server logs:

  1. On your self-hosted GitLab, can you check the webhook logs in /var/log/gitlab/gitlab-rails/production.log? Look for entries right after your follow-up push.

  2. Alternatively, in the webhook settings there’s a “Test” button. Try Test > Push events. Do you get a response from Cursor?

  3. You can try creating a new MR with a single commit. Does the first review trigger? This helps confirm if the issue is only with follow-up pushes or with all new MRs.

I tried with the 2nd instruction, and I got the response 200, so the connection is good.

image

  1. I am pretty sure that the new MRs will always trigger the review, but follow-up pushes does not. We have several new MRs triggered the review, although our devs resolved the issues and pushed new commit, and even merged the MR, there is no updates on the cursor admin panel. Here is the screenshot you can check.

Thanks for the info. The webhook is connected correctly (HTTP 200 on test), and the issue is confirmed: new MRs trigger reviews, but follow-up pushes are not processed.

This seems like a bug in handling webhook events from self-hosted GitLab after the first review.

Is there an estimated fix time for this issue?

I’ve already shared this with the team, and they’re looking into it. Unfortunately, I can’t provide an exact ETA for a fix yet.

Got it, please keep me updated, thanks. Looking forward for the fix.

1 week has passed, is there any updates on this issue?

Thanks for your patience. The team is aware of the issue and is working on a fix. Unfortunately, I can’t share an exact timeline yet. I’ll update the thread as soon as there’s progress.

Totally understand, looking forward to your updates.

1 Like

Another week passed, any updates on this issue?

Hi all, our team is still looking into this!

@Yuan_Gao are you able to provide another recent example request ID by any chance?

Here you are: Bugbot request id: serverGenReqId_ba485da8-48f0-4736-b00b-d9bb5fc2170b

The request of the request id provided above does work correctly, so this bug is fixed?

Interesting - glad to hear!

We haven’t been able to reproduce the reported issue with our own Gitlab setup, so if everything is working well on your end, this is likely resolved. Let us know if this crops up again and we’ll take another look.

That is very frustrating when the bot locks up after the initial run. Since the initial connection and review worked, I believe the issue is a server-side state lock related to that specific Merge Request ID after the first successful execution.

I suggest checking two things: first, confirm in your self-hosted GitLab configuration that the push event webhook is successfully firing upon subsequent commits. Second, if the webhooks look correct, the BugBot server logs associated with that request ID should show where the process is exiting or why it is not picking up the queue item. It is likely getting interrupted due to a failed state check.

Hope this helps!