By popular demand, this patch lets folks manually label PRs with the
label and make it stick, even if auto-labeler wouldn't label it itself
(presumably because the author didn't mark their commits according to
automation policy).
Assisted-by: Codex gpt-5.6-sol medium
Discussed in AGENTS.md debate PR
Assisted by https://regex101.com/, my brain, and love for humanity and free software
Signed-off-by: Ethan Carter Edwards <ethan@ethancedwards.com>
`bot.js` already pulls reviews via GraphQL for the approval-count
labels. Move that fetch above the `handleMerge` call, add `commit
{ oid }` to the query, and pass the result through. `handleMerge` no
longer issues its own `listReviews` REST call.
While here, the `approvals` set is now derived from the same reviews
data. Two upsides:
- A reviewer who approved and was later dismissed no longer counts
towards the approval check; their review now surfaces with state
DISMISSED instead of leaving a stale `reviewed` event behind.
- `events` is no longer needed by `runChecklist` (it was only feeding
the approvals filter); the parameter is dropped from both call
sites. `handleMerge` still uses `events` for tracking the latest
push and merge-command comments.
Also drop the redundant `user` truthy checks (and the stale "some
users have been deleted" comment) in `runChecklist`. The GraphQL
query uses `author { ... on User { login id } }` and bot.js then
filters via `r.user?.login`, so by the time reviews reach
`runChecklist` every entry already has a populated `user`. Verified
by querying real PRs: bots surface as `{__typename: "Bot"}` (no
`login`/`id`, filtered out by bot.js), and deleted accounts surface
as the "ghost" user (login `"ghost"`, id `10137`), which passes the
filter but matches no committer - harmless for both checks.
Assisted-by: claude-code with claude-opus-4-8[1m]-high
New feature: revoke review requests from a reviewer when the reason they
were review-requested in the first place is gone (e.g. force-pushed),
while preserving reviewers who voluntarily engaged.
Checks all PR commit messages for version update pattern
like 'packagename: X.Y.Z -> A.B.C'.
Matches: 1.2.3, 0-unstable-2024-01-15, 1.3rc1, alpha, unstable
Apparently any effects from this change haven’t shown up noticeably
in GitHub’s metrics, to the point where they’re not sure if it
was taking effect on the backend. Our contact is going to look at
getting something into the API response to help debug whether it’s
actually working or not, but agreed that we should just revert for
now. Since they have apparently reduced replica sync issues further
through other changes on their end, there shouldn’t be any urgent
need to make any changes here anyway.
This reverts commit 18b30c8ce1.
When a maintainer deletes their GitHub account, the bot would crash with a 404 error when trying to fetch their user info via `/user/{id}`.
This caused the scheduled bot workflow to fail repeatedly until manual intervention (e.g., closing/reopening the affected PR to clear the requested reviewer).
Fix by returning null from getUser() for 404 responses and filtering out null users when building the reviewers list.
This is an experiment and can be reverted a few days from now; if
the results are positive on GitHub’s end, then we may want to make
the merge conflict checks run less frequently than the rest of the
labelling tasks.
I fixed this for the maintainer maps, but the artifact that was causing the
particular issue that prompted me to try to fix it was actually the
pagination cursor. So fix that too.
Related: #464046.
This reverts commit 402b41c125.
GitHub' API repeatedly returns wrong data which causes closed PRs when
the changes had not been merged, yet.
We have closed a bit more than 100 PRs overall, most of them initially -
the feature is not really that important overall.
It makes not much sense to run all the checks for PRs when we can
already tell they are stale beforehand. In particular this should avoid
creating ~3.3k temporary merge commits every day, for PRs that surely
won't have had any change.
The number of merge commits *could* play a role in the growing size of
the fork network. We'll have GitHub look into the metrics before and
after this change to see whether that is any improvement.
This was introduced as part of the hotfix PR to avoid hitting API rate
limits - but the condition was wrong. It was supposed to trigger in all
PR contexts, not only for the Test workflow.
We only recently introduced the owners.txt file to the comparison
artifact, so once the bot runs on a schedule it will it older artifacts
very quickly - and then can't find the owners file.
We can fallback to an empty owners list in this case, because an older
artifact also means an older workflow run previously, so this will have
pinged owners already.
This migrates the bash code to request reviewers to github-script. This
will allow multiple nice improvements later on, but at this stage it's
mostly a reduction in code and complexity.
We technically counted bot approvals and approvals by deleted users for
the approval labels as well. The former don't exist, yet, but if they
were, I don't think we'd count them. The latter should arguably *not* be
counted, because we can't tell anymore *who* approved, so we can't put
any weight on it as reviewers.
This simplifies the logic, too.
