[Dev][CI] Extra labels are confusing

Describe the bug, including details regarding any error messages, version, and platform.

When you add an Extra label for CI, sometimes it gets later removed by a bot. It's not obvious why or when this happens.

Component(s)

Developer Tools, Continuous Integration

@kou

@kou

Could you share a sample PR that the problem happened?

#50125 (comment) ?

@raulcd

@raulcd It seems that

arrow/dev/archery/archery/bot.py

Lines 214 to 216 in 9265a71

    
           def set_state(self, state): 
        
               """Sets the State label to the PR.""" 
        
               self.pull.add_to_labels(state.value)

that uses https://docs.github.com/en/rest/issues/labels?apiVersion=2026-03-10#add-labels-to-an-issue may remove existing labels when multiple label add requires are processed at once. (I think that it's a bug of GitHub...)

Are we using the "awaiting ..." labels? If we don't use the "awaiting ..." labels, how about removing them? If we remove them, we can avoid this timing sensitive problem.

I've just used the "Add labels to an issue" API as instructed here to add the Java label to this issue, it didn't remove other labels.

Note that "Add labels to an issue" uses POST while "Set labels for an issue" uses PUT.

The GHA-based issue bot uses setLabels instead of addLabels or removeLabel and that might be the issue.

#50125 (comment) is a problem case, right?

It's a PR. Our issue bot isn't executed for PR:

arrow/.github/workflows/issue_bot.yml

Lines 20 to 23 in 4e25461

#50125 (comment) (you added CI: Extra: C++ and CI: Extra: R) happened on 2026-06-08T14:02:30Z and
#50125 (comment) (Archery added awaiting review and removed CI: Extra: C++) happened on 2026-06-08T14:02:40Z.

I think that this timeline is related...

#50125 (comment) is a problem case, right?

Yes. It happens to me often actually.

It's a PR. Our issue bot isn't executed for PR:

Hmm. I see.

I think that this timeline is related...

It is. But it's weird because add_labels is not supposed to remove existing labels...

I think the problem is that we're using https://github.com/actions/labeler for PRs which uses setLabels.

Another project had the same problem with it: flutter/flutter#128440

Also see actions/labeler#763 . They claim the issue is fixed but I don't understand how.

@kou

Ok, the AI upstream said this:

In Arrow's labeler.yml, CI: Extra: C++ is configured as a labeler-managed label. PR #50125 modified Docker and compose files, which do not match the configured rules for that label. As a result, when labeler ran with sync-labels: true, it removed CI: Extra: C++ as expected.

It looks like the AI is right. @kou Do we need to add more paths to labeler.yml? Or should we perhaps make the "Extra" labels entirely manual (non-automated)?

I'd personally remove them from the labeler.yml and make the extra labels completely manual. I don't think trying to be smart adding them automatically is that useful and I personally trigger them when I think a PR requires those jobs to be executed. Similar to how we trigger crossbow jobs basically.

I think that the answer isn't right.

The labelar job is executed when PR is opened or synchronized (pushed):

arrow/.github/workflows/dev_pr.yml

Lines 20 to 30 in cffe22d

    
           on: 
        
             # TODO: Enable this when eps1lon/actions-label-merge-conflict is available. 
        
             # push: 
        
             #   branches: 
        
             #     - master 
        
             #     - main 
        
             pull_request_target: 
        
               types: 
        
                 - opened 
        
                 - edited 
        
                 - synchronize

arrow/.github/workflows/dev_pr.yml

Lines 86 to 94 in cffe22d

    
                 - name: Assign GitHub labels 
        
                   if: | 
        
                     (github.event.action == 'opened' || 
        
                      github.event.action == 'synchronize') 
        
                   uses: actions/labeler@v6 
        
                   with: 
        
                     repo-token: ${{ secrets.GITHUB_TOKEN }} 
        
                     configuration-path: .github/workflows/dev_pr/labeler.yml 
        
                     sync-labels: true

The CI: Extra: C++ label isn't removed after you set it manually again #50125 (comment) and then you pushed to the branch again #50125 (comment) .

If the answer is right, the CI: Extra: C++ label should be removed when you pushed to the branch again. But the CI: Extra: C++ label wasn't removed.

apacheGH-50133: [CI] Make extra labels entirely manual

GH-50133: [CI] Make extra labels entirely manual on PRs (#50256)

Issue resolved by pull request 50256
#50256

Navigation Menu

Repository navigation

[Dev][CI] Extra labels are confusing #50133

Describe the bug, including details regarding any error messages, version, and platform.

Component(s)

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

	def set_state(self, state):
	"""Sets the State label to the PR."""
	self.pull.add_to_labels(state.value)

	on:
	# TODO: Enable this when eps1lon/actions-label-merge-conflict is available.
	# push:
	# branches:
	# - master
	# - main
	pull_request_target:
	types:
	- opened
	- edited
	- synchronize

	- name: Assign GitHub labels
	if: \|
	(github.event.action == 'opened' \|\|
	github.event.action == 'synchronize')
	uses: actions/labeler@v6
	with:
	repo-token: ${{ secrets.GITHUB_TOKEN }}
	configuration-path: .github/workflows/dev_pr/labeler.yml
	sync-labels: true

[Dev][CI] Extra labels are confusing #50133

Description

Describe the bug, including details regarding any error messages, version, and platform.

Component(s)

Activity

pitrou commented on Jun 9, 2026

kou commented on Jun 9, 2026

kou commented on Jun 10, 2026

kou commented on Jun 10, 2026

pitrou commented on Jun 10, 2026

pitrou commented on Jun 10, 2026

pitrou commented on Jun 10, 2026

kou commented on Jun 10, 2026

pitrou commented on Jun 10, 2026

pitrou commented on Jun 10, 2026

pitrou commented on Jun 10, 2026

pitrou commented on Jun 25, 2026

raulcd commented on Jun 25, 2026

kou commented on Jun 25, 2026

raulcd commented on Jun 30, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions