Sudden issues with gitlab PR metadata, error on list topics

[mplanchard]

At some point yesterday, my status buffer stopped showing my custom sections of PRs assigned to me and PRs on which my review has been requested.

GitLab had some scheduled maintenance ongoing related to the OAuth API, so I figured it might be related. In the meantime, I updated emacs itself to 30.1 and all of my emacs packages, including magit.

This morning, I'm still seeing the same behavior, so I figured I'd ask about it.

Essentially, all of the assignee and reviewer data on all PRs seems to be gone. There is also no autocomplete for reviewer/assignee usernames when I press C-c C-e to edit those fields. Typing out a full username and pressing enter does not lead to any errors, but it also does not update the state on GitLab.

If I open the forge transient menu with @ and type l t for list topics, the following error is displayed in the Messages buffer:

EmacSQL had an unhandled condition: nil, nil, nil, ("SQL logic error" "no such table: discussion_assignee" 1 1)

I don't see any other errors when pulling PRs, attempting to assign users, etc.

What I've tried:

• removing the repository using forge-remove-repository and re-adding it, but it winds up in the same state.
• deleting the entire forge-database.sqlite file from my emacs directory and reinitializing everything, but that also leads to the same state.
• making a new personal access token with more permissions. The previous one just had api, the new one has api, read_api, read_user, read_repository, write_repository

Versions:

package	version
forge	0.5.0 9db4d38
compat	30.1.9999 (builtin)
closql	2.2.1 dc7924c
emacsql	4.3.0 5470ada
ghub	4.3.0 1fbce53
let-alist	1.0.6 (builtin)
llama	0.6.2 48e5bc4
magit	4.3.2 bf58615
markdown-mode	2.8-alpha 258313e
seq	2.24 (builtin)
transient	0.8.7 afc88b2
yaml	1.2.0 f99ef76

I don't think it's related because nothing about them has changed, but these are my custom sections:

(use-package forge :ensure t
  :after magit
  :config
  ;; re-add support for assigned & review requests at top level of status
  ;; buffer without additional filtering. See https://github.com/magit/forge/issues/676
  (defun my/forge-insert-assigned-pullreqs ()
    "Insert assigned pullreqs to the magit status buffer."
    (forge-insert-topics 'assigned-pullreqs "Assigned pull requests"
      (lambda (repo)
        (and-let* ((me (ghub--username repo)))
          (forge--topics-spec :type 'pullreq :active t :assignee me)))))
  (defun my/forge-insert-review-requests ()
    "Insert requested reviews to the magit status buffer."
    (forge-insert-topics 'reviewer-pullreqs "Review requests"
      (lambda (repo)
        (and-let* ((me (ghub--username repo)))
          (forge--topics-spec :type 'pullreq :active t :reviewer me)))))
  (magit-add-section-hook 'magit-status-sections-hook
                          #'my/forge-insert-assigned-pullreqs
                          #'forge-insert-pullreqs)
  (magit-add-section-hook 'magit-status-sections-hook
                          #'my/forge-insert-review-requests
                          #'my/forge-insert-assigned-pullreqs))

[mplanchard]

The forge-database.sqlite file currently contains the following tables:

Table Name	Number of Rows
assignee	0
discussion	0
discussion_category	0
discussion_label	0
discussion_mark	0
discussion_post	0
discussion_reply	0
fork	0
issue	1
issue_assignee	0
issue_label	0
issue_mark	0
issue_post	8
label	0
mark	0
milestone	0
notification	0
pullreq	144
pullreq_assignee	0
pullreq_label	0
pullreq_mark	0
pullreq_post	1954
pullreq_review_request	0
repository	1
revnote	0

So it does indeed look like discussion_assignee is absent, and also that there are no rows for pullreq_asignee or pullreq_review_request

[tarsius]

I plan on working on Gitlab support next month and will look at this then.

[tarsius]

Please check whether b0c0de7 fixed this issue.

[mplanchard]

On current main (0f5f2a3), it seems that the list-topics error is no longer occurring.

However, again after having removing and re-adding the forge repository with forge-remove-repository, MRs don't have any asignee/reviewer data, Trying to assign an asignee or reviewer actually removes any existing asignee/reviewer in GL, but does not assign anyone.

[mplanchard]

The sqlite database still appears to have no data for asignees/reviewers, as well:

assignee                              0 
discussion                            0 
discussion_category                   0 
discussion_label                      0 
discussion_mark                       0 
discussion_post                       0 
discussion_reply                      0 
fork                                  0 
issue                                 1 
issue_assignee                        0 
issue_label                           0 
issue_mark                            0 
issue_post                            8 
label                                 0 
mark                                  0 
milestone                             0 
notification                          0 
pullreq                             240 
pullreq_assignee                      0 
pullreq_label                         0 
pullreq_mark                          0 
pullreq_post                       3385 
pullreq_review_request                0 
repository                            1 
revnote                               0 

[mplanchard]

I expect it's some change in their API, since I the issue first started happening before I ever updated forge. Not getting any errors anywhere, but happy to provide any debugging output that might be helpful.

[tarsius]

I expect to focus on Forge next month. Edit: Ah, I see I already set that two months ago. Well, I still plan for that to happen this next month. (And while I will actually focus on Forge today and maybe the next few days too, it won't be this particular issue.)

[tarsius]

I've just tried to reproduce this once more but still can't. I used a fresh local database and then added https://gitlab.com/python-mode-devs/python-mode. Can you reproduce the issue with that repository? I made sure that the assignee is shown when there is one, e.g., for issue 154.

What repository did you use? Maybe I can reproduce with that repository too.

I am closing this for now, but if you can still reproduce and provide steps that would allow me to reproduce, then I would of course be interested in fixing that.