morluto/rea

Fork 0

mirror of https://github.com/morluto/rea.git synced 2026-11-05 06:45:22 +00:00

[GH-ISSUE #3432] [Bug] PR readiness can never succeed for PRs from forks: the head-association lookup queries the base repository #1278

New issue

Closed

opened 2026-10-16 21:03:41 +00:00 by bsakaguchi · 4 comments

bsakaguchi commented

2026-10-16 21:03:41 +00:00

Owner

Originally created by @pablodiazjorge on GitHub (Oct 14, 2026).
Original GitHub issue: https://github.com/morluto/rea/issues/3432

Summary

scripts/ci/pr-readiness.mjs (from #3302) confirms the head association with listPullRequestsAssociatedWithCommit against context.repo, the base repository. For a commit that exists only in a fork, that endpoint returns an empty list, so every fork PR whose declaration passes evaluateReadiness ends at pending with "GitHub has not confirmed this PR's head association; rerun PR readiness." A rerun cannot change the result, because the lookup is still made against the base repository.

Evidence

The same endpoint, for the head commits of three fork PRs, queried against each repository on 2026-10-14 around 10:50 UTC:

PR	head	`repos/morluto/rea/commits/<sha>/pulls`	`repos/<fork>/rea/commits/<sha>/pulls`
#3266	`f73ad29f2`	`[]`	`#3266` (base `morluto/rea`)
#3159	`c68d96c94`	`[]`	`#3159` (base `morluto/rea`)
#2148	`46ad54d2e`	`[]`	`#2148` (base `morluto/rea`)

The workflow runs for these three PRs (38548084743, 38548089515, 38548095313) all concluded success, and each left the PR readiness status at pending with the head-association message. Across the 40 most recent open PRs, no PR readiness status is success. Each one is failure for a declaration problem, pending for a draft, or pending with the head-association message, and that last group includes non-draft fork PRs such as #3298, #3305, #3319, #3409, #3412, #3418 and #3422.

Expected

A fork PR with a valid Author readiness section reaches success.

Possible fix

Run the association lookup against the head repository when it differs from the base (pull.head.repo.owner.login and pull.head.repo.name). That listing returns the PR into morluto/rea, as the table shows. Alternatively, re-fetch the PR with pulls.get just before publishing and compare head.sha, which covers the race the guard seems to target without depending on the commit index. The distinct-head check would still see other open PRs that share the commit from the same fork. It would not see PRs from a second fork, if that case matters.

Not verified: a patched run in this repository, since the workflow runs on pull_request_target from the default branch.

Originally created by @pablodiazjorge on GitHub (Oct 14, 2026). Original GitHub issue: https://github.com/morluto/rea/issues/3432 ### Summary `scripts/ci/pr-readiness.mjs` (from #3302) confirms the head association with `listPullRequestsAssociatedWithCommit` against `context.repo`, the base repository. For a commit that exists only in a fork, that endpoint returns an empty list, so every fork PR whose declaration passes `evaluateReadiness` ends at `pending` with "GitHub has not confirmed this PR's head association; rerun PR readiness." A rerun cannot change the result, because the lookup is still made against the base repository. ### Evidence The same endpoint, for the head commits of three fork PRs, queried against each repository on 2026-10-14 around 10:50 UTC: | PR | head | `repos/morluto/rea/commits/<sha>/pulls` | `repos/<fork>/rea/commits/<sha>/pulls` | |---|---|---|---| | #3266 | `f73ad29f2` | `[]` | `#3266` (base `morluto/rea`) | | #3159 | `c68d96c94` | `[]` | `#3159` (base `morluto/rea`) | | #2148 | `46ad54d2e` | `[]` | `#2148` (base `morluto/rea`) | The workflow runs for these three PRs (38548084743, 38548089515, 38548095313) all concluded `success`, and each left the `PR readiness` status at `pending` with the head-association message. Across the 40 most recent open PRs, no `PR readiness` status is `success`. Each one is `failure` for a declaration problem, `pending` for a draft, or `pending` with the head-association message, and that last group includes non-draft fork PRs such as #3298, #3305, #3319, #3409, #3412, #3418 and #3422. ### Expected A fork PR with a valid Author readiness section reaches `success`. ### Possible fix Run the association lookup against the head repository when it differs from the base (`pull.head.repo.owner.login` and `pull.head.repo.name`). That listing returns the PR into `morluto/rea`, as the table shows. Alternatively, re-fetch the PR with `pulls.get` just before publishing and compare `head.sha`, which covers the race the guard seems to target without depending on the commit index. The distinct-head check would still see other open PRs that share the commit from the same fork. It would not see PRs from a second fork, if that case matters. Not verified: a patched run in this repository, since the workflow runs on `pull_request_target` from the default branch.

bsakaguchi closed this issue

2026-10-16 21:03:41 +00:00

bsakaguchi commented

2026-10-16 21:03:42 +00:00

Author

Owner

@pablodiazjorge commented on GitHub (Oct 14, 2026):

The PR readiness gate was retired at 18:52 UTC ("CI gate retired; author readiness was not evaluated"), and fork PRs now pass it, so this no longer blocks anything. Closing; happy to reopen if the gate comes back in a form that still queries the base repository.

@pablodiazjorge commented on GitHub (Oct 14, 2026): The PR readiness gate was retired at 18:52 UTC ("CI gate retired; author readiness was not evaluated"), and fork PRs now pass it, so this no longer blocks anything. Closing; happy to reopen if the gate comes back in a form that still queries the base repository.

bsakaguchi referenced this issue

2026-10-17 21:09:56 +00:00

[PR #1278] [MERGED] docs(ghidra): explain supported heap and CPU controls #2760

bsakaguchi commented

2026-10-31 00:30:25 +00:00

Author

Owner

@pablodiazjorge commented on GitHub (Oct 19, 2026):

Reopening as I said I would: PR readiness is active again since 2026-10-18 22:46 UTC with the same script as #3302 (scripts/ci/pr-readiness.mjs is unchanged), and the head-association lookup still queries the base repository.

For PRs from forks, listPullRequestsAssociatedWithCommit on morluto/rea returns no pull request. The heads of #4903 (e67abba99), #4103 (4a1103e8a) and #2148 (1f4da4d9c) all return [], so their declarations pass and the check still ends in pending: GitHub has not confirmed this PR's head association; rerun PR readiness. A rerun gives the same answer. Among the 60 most recent open PRs, 18 from forks are in that state and none has reached success; the rest are drafts or have incomplete declarations.

@peter-research's #3461, which looks the association up in the head repository, was closed on 15 Oct only because the gate had been retired. It touches scripts/ci/pr-readiness.mjs, which hasn't changed since, so it may still apply.

@pablodiazjorge commented on GitHub (Oct 19, 2026): Reopening as I said I would: PR readiness is active again since 2026-10-18 22:46 UTC with the same script as #3302 (`scripts/ci/pr-readiness.mjs` is unchanged), and the head-association lookup still queries the base repository. For PRs from forks, `listPullRequestsAssociatedWithCommit` on morluto/rea returns no pull request. The heads of #4903 (`e67abba99`), #4103 (`4a1103e8a`) and #2148 (`1f4da4d9c`) all return `[]`, so their declarations pass and the check still ends in `pending: GitHub has not confirmed this PR's head association; rerun PR readiness.` A rerun gives the same answer. Among the 60 most recent open PRs, 18 from forks are in that state and none has reached `success`; the rest are drafts or have incomplete declarations. @peter-research's #3461, which looks the association up in the head repository, was closed on 15 Oct only because the gate had been retired. It touches `scripts/ci/pr-readiness.mjs`, which hasn't changed since, so it may still apply.

bsakaguchi commented

2026-10-31 00:30:25 +00:00

Author

Owner

@BaaRai commented on GitHub (Oct 19, 2026):

Same state here on #1922, #5023 and #5025. After filling the Author readiness declarations (owner plus the three checked statements), evaluateReadiness itself passes and the check still ends at pending: GitHub has not confirmed this PR's head association, because listPullRequestsAssociatedWithCommit on morluto/rea returns [] for each head (8944740b5, 5c2a90a3e, fd60316d9, and the updated tips after merging main).

Concrete mergeability impact: #1922 has every other lane green (source tests, all five shards, portable, static checks, real lanes), so this pending status is the only thing blocking it.

Happy to help validate a fix that queries the head repository, for example by re-running the readiness workflow on these three PRs once it lands.

@BaaRai commented on GitHub (Oct 19, 2026): Same state here on #1922, #5023 and #5025. After filling the Author readiness declarations (owner plus the three checked statements), `evaluateReadiness` itself passes and the check still ends at `pending: GitHub has not confirmed this PR's head association`, because `listPullRequestsAssociatedWithCommit` on morluto/rea returns `[]` for each head (`8944740b5`, `5c2a90a3e`, `fd60316d9`, and the updated tips after merging main). Concrete mergeability impact: #1922 has every other lane green (source tests, all five shards, portable, static checks, real lanes), so this pending status is the only thing blocking it. Happy to help validate a fix that queries the head repository, for example by re-running the readiness workflow on these three PRs once it lands.

bsakaguchi commented

2026-10-31 00:30:25 +00:00

Author

Owner

@pablodiazjorge commented on GitHub (Oct 20, 2026):

Fixed by #5316 (merged 2026-10-19 23:19 UTC, thanks @gensokyo-dev): when the association lookup returns no pull request, readiness now checks the live open PR heads. #4903, #4103 and #2148 report PR readiness success since then. Closing.

@pablodiazjorge commented on GitHub (Oct 20, 2026): Fixed by #5316 (merged 2026-10-19 23:19 UTC, thanks @gensokyo-dev): when the association lookup returns no pull request, readiness now checks the live open PR heads. #4903, #4103 and #2148 report PR readiness success since then. Closing.

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

morluto/rea#1278

No description provided.

Rows
Columns