pr-review posting: REQUEST_CHANGES event and PR-body fallback list #1528

Closed
opened 2026-10-27 13:40:45 +00:00 by sam · 2 comments
Owner

docs/prds/pr-review-agent.md phase 9 says a blocking finding requests changes, and docs/prds/inline-review.md says a whole-file finding (no line) falls back to a PR-body list. Neither is wired for the pr-review pipeline yet (internal/domain/workflow/prreview_stages.go, postPRReview):

  1. workflow.Forger.CreateReviewComments always posts a COMMENT-state review. Its GitHub implementation (internal/forge/github.go) doesn't submit a review object at all -- it posts per-comment via PullRequests.CreateComment -- so REQUEST_CHANGES needs a Forger/forge interface change plus a GitHub implementation change from per-comment calls to a submitted review with an event. Gitea's CreateReviewComments already submits one review object and could take an event parameter more directly.
  2. A finding with no line (LineStart <= 0) is currently dropped in postPRReview rather than listed anywhere, unlike the legacy single-reviewer path's PR-body fallback list (docs/prds/inline-review.md decision 3).

prreview.ReviewEventFor and Finding.Blocking already exist and are correctly set by the merge gate (archie-core-afbk.5); only the forge-posting side needs this follow-up.

Notes

Core side done 94e6354d; extension side in 8i87z


Imported from beads archie-core-q832 · GitHub #1349 · type task · priority P2 · created 2026-09-28

docs/prds/pr-review-agent.md phase 9 says a blocking finding requests changes, and docs/prds/inline-review.md says a whole-file finding (no line) falls back to a PR-body list. Neither is wired for the pr-review pipeline yet (internal/domain/workflow/prreview_stages.go, postPRReview): 1. workflow.Forger.CreateReviewComments always posts a COMMENT-state review. Its GitHub implementation (internal/forge/github.go) doesn't submit a review object at all -- it posts per-comment via PullRequests.CreateComment -- so REQUEST_CHANGES needs a Forger/forge interface change plus a GitHub implementation change from per-comment calls to a submitted review with an event. Gitea's CreateReviewComments already submits one review object and could take an event parameter more directly. 2. A finding with no line (LineStart <= 0) is currently dropped in postPRReview rather than listed anywhere, unlike the legacy single-reviewer path's PR-body fallback list (docs/prds/inline-review.md decision 3). prreview.ReviewEventFor and Finding.Blocking already exist and are correctly set by the merge gate (archie-core-afbk.5); only the forge-posting side needs this follow-up. ### Notes Core side done 94e6354d; extension side in 8i87z --- _Imported from beads `archie-core-q832` · GitHub #1349 · type task · priority P2 · created 2026-09-28_
sam added the mediumtask labels 2026-10-27 13:40:45 +00:00
sam self-assigned this 2026-10-27 13:40:45 +00:00
sam added a new dependency 2026-10-27 13:42:26 +00:00
Author
Owner

Triaged: no code change is needed for this. Both requested pieces are already wired in this repo, so no change is needed here.

  1. Review event. workflow.Review has Event/Body (internal/domain/workflow/workflow.go:49-54), forge.ReviewSubmission has the same fields (internal/forge/forge.go:186-194), the NATS request carries them (internal/forgerpc/forgerpc.go:125-135, client at 477-491, server handler at 323-350), and the extension proto declares string event = 4 and string body = 5 (proto/forge/v1/forge.proto:218-227), which forgeext forwards (internal/infrastructure/forgeext/forgeext.go:317-326). forgeReview/reviewEvent in prreview_stages.go:702-731 set REQUEST_CHANGES when any finding blocks, falling back to COMMENT for the task's own PR (GitHub refuses REQUEST_CHANGES from the PR author). Tests exist: TestForgeReviewEvent covers all three cases.

  2. No-line fallback. forgeReview sends a finding inline only when prreview.MapLine(...) is not LineOutside; otherwise it appends "- file:line: body" to the review body, using just the path when LineStart<=0. MapLine returns LineOutside for any line<=0 (prreview/diff.go:682-685), and the refusal retry re-renders with files=nil, putting every finding in the body. TestForgeReviewPlacesOnlyDiffLinesInline has a "no line" case asserting the finding is not dropped.

CHANGELOG.md:94 already lists this as shipped: "Pull request reviews submit REQUEST_CHANGES on blocking findings, list findings outside the diff in the review body, and repost without inline comments when the forge refuses their placement."

The GitHub implementation the issue names (internal/forge/github.go, per-comment PullRequests.CreateComment) is not in this worktree: internal/forge holds only forge.go/noop.go/access.go/webhook, and go.mod has no go-github or go-gitea dependency. That adapter lives in the extension, which the issue notes already track separately as 8i87z. So the only residual work is outside this repository and cannot be judged from here.

Verified by reading code and tests; I did not run the test suite (no execution tool in this read-only mission).

Triaged: no code change is needed for this. Both requested pieces are already wired in this repo, so no change is needed here. 1. Review event. `workflow.Review` has `Event`/`Body` (internal/domain/workflow/workflow.go:49-54), `forge.ReviewSubmission` has the same fields (internal/forge/forge.go:186-194), the NATS request carries them (internal/forgerpc/forgerpc.go:125-135, client at 477-491, server handler at 323-350), and the extension proto declares `string event = 4` and `string body = 5` (proto/forge/v1/forge.proto:218-227), which forgeext forwards (internal/infrastructure/forgeext/forgeext.go:317-326). `forgeReview`/`reviewEvent` in prreview_stages.go:702-731 set REQUEST_CHANGES when any finding blocks, falling back to COMMENT for the task's own PR (GitHub refuses REQUEST_CHANGES from the PR author). Tests exist: TestForgeReviewEvent covers all three cases. 2. No-line fallback. `forgeReview` sends a finding inline only when `prreview.MapLine(...)` is not LineOutside; otherwise it appends "- `file`:`line`: body" to the review body, using just the path when LineStart<=0. MapLine returns LineOutside for any line<=0 (prreview/diff.go:682-685), and the refusal retry re-renders with `files=nil`, putting every finding in the body. TestForgeReviewPlacesOnlyDiffLinesInline has a "no line" case asserting the finding is not dropped. CHANGELOG.md:94 already lists this as shipped: "Pull request reviews submit REQUEST_CHANGES on blocking findings, list findings outside the diff in the review body, and repost without inline comments when the forge refuses their placement." The GitHub implementation the issue names (internal/forge/github.go, per-comment PullRequests.CreateComment) is not in this worktree: `internal/forge` holds only forge.go/noop.go/access.go/webhook, and go.mod has no go-github or go-gitea dependency. That adapter lives in the extension, which the issue notes already track separately as 8i87z. So the only residual work is outside this repository and cannot be judged from here. Verified by reading code and tests; I did not run the test suite (no execution tool in this read-only mission).
Author
Owner

Closed: goes against the product plan in epic #1977. Archie isn't a coding agent; coding is a side effect. PR review stops being a core product line, and the first story is proactive day-2 operations. Forge, repository and review behaviour moves behind a forge connector (#1657, #1978).

Closed: goes against the product plan in epic #1977. Archie isn't a coding agent; coding is a side effect. PR review stops being a core product line, and the first story is proactive day-2 operations. Forge, repository and review behaviour moves behind a forge connector (#1657, #1978).
sam removed a dependency 2026-11-14 14:45:28 +00:00
sam closed this issue 2026-11-14 14:45:28 +00:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: sam/archie#1528