Repository navigation
Conversation
…rs pane errorItems and warningItems were gated on `cleared`, which is `!track.active` -- and `active` means "inside a chapter that has not ended" (useTaskEvents.ts:646), not "a run exists". So the moment a run finished and endTime was set, both lists were emptied one line before render and the pane displayed "No errors or warnings" about arrays it had just discarded. A run that failed thirty seconds ago reported having raised nothing. The pane's own empty-state copy stated the assumption -- "appear here while it runs or when replaying a recorded run" -- which is a correct description for a streaming window and wrong for a status snapshot. Updated with the fix. Diagnostics raised when a node is configured rather than run have no chapter at all, so the same gate discarded them permanently. Outside a track the panes now fall back to the task's own status. Replay still blanks, because there the question genuinely is "what was true at this position". The accumulating panes (tokens, StatusPane) stay on `cleared`, where that gate is correct. Refs rocketride-org#1832
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: 📒 Files selected for processing (1)
📝 WalkthroughWalkthroughSourcePanel now preserves errors and warnings from completed runs by using live task status outside replay mode. Replay remains position-scoped. The empty-state text now reflects completed-run diagnostics. ChangesDiagnostics retention
Estimated code review effort: 2 (Simple) | ~10 minutes Suggested reviewers: 🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches 💡 1🛠️ Fix failing CI checks 💡
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
🤖 Internal: Discord sync markerAuto-managed by the Discord notification workflow. Stores the linked Discord message ID and forum thread ID. Do not edit or delete. |
|
The For comparison, macOS (ARM64) is green on the other PRs I have open right now, including larger ones: #1815, #1819, #1821, #1823, #1825, #1827. Locally: I do not have rerun rights on this repo — could someone re-run the failed job when convenient? Happy to push an empty commit instead if you would rather, though that would dismiss any review already in flight. |
|
Following up on my earlier comment with the actual cause, now that the run's logs are available — it is a runner network failure, not the change. The macOS (ARM64) runner lost connectivity to the npm registry during dependency install: Sixteen packages, This PR changes one TypeScript file — 13 lines in A re-run should clear it. I do not have rerun rights here — happy to push an empty commit instead if you would prefer, though that would dismiss any review in flight. |
|
@coderabbitai review |
✅ Action performedReview finished.
|
|
Still worth doing, and the diagnosis holds on today's develop: The blocker is that the file moved. #1843 (488d076, 2026-08-07) replaced |
Review: #1833 fix(shared-ui): keep a finished run's diagnostics visible in the Errors paneVerdict: request changes, but only because the branch no longer merges and CI is red and stale. The reasoning and the three-line fix hold up under everything I threw at them, and the bug is still live on Metrics
Invariants checkedDerived from AGENTS.md, CODEOWNERS, the CI workflows, and the code itself.
Findings[medium] The branch no longer merges: the file moved out from under it Refutation attempted: I first checked whether git could resolve it by rename detection, and with detection on and the old empty-state copy is still at :591. So the fix is still needed and still applies. [medium] Last CI run is red, and its logs are gone [low] The new empty-state copy is wrong on the DEPLOY tab [low] The fix does not survive a reconnect [low] No test covers the fixed behavior Refutation log
I did not attach a screenshot: reproducing this needs the VS Code extension running against a live engine with a failing pipeline, which I could not drive here. The behavior is established from the source path above instead. Review ran in an isolated worktree at commit @jmaionchi @Rod-Christensen @stepmikhaylov the fix itself looks right to me and the bug is still present on develop at |
joshuadarron
left a comment
There was a problem hiding this comment.
Verdict: Request changes
The PR stops the Errors pane from emptying when a run ends: outside a live track, it falls back to the task's live status instead of []. The bug is real and still present on develop. The PR can't merge as it stands: the file has moved, CI is failing, and the fix pulls from a second status source and has no test.
Blocking
- Merge conflicts with the base branch. Rebase or merge the base branch and resolve them.
- CI checks failing: Build / macOS (ARM64), CI OK
- UI files changed with no screenshots in the PR body. Add before and after screenshots. Files:
- packages/shared-ui/src/modules/project/components/SourcePanel.tsx
SourcePanel.tsxno longer exists at this path ondevelop. #1843 moved it toapps/shared/src/modules/project/components/SourcePanel.tsx, and the bug is still there at:534. Rebase and apply the fix to the new path.SourcePanel.tsx:544Replace the nested ternary withifblocks (code-taste: readability). Better still, extract the selection into a small exported pure function, for exampleselectDiagnosticsStatus({ isCleared, isReplay, paneStatus }).SourcePanel.tsx:544UsepaneStatushere, notliveTaskStatus. The block comment at:370-376says the snapshot panes read from the stream in both modes so that they have "one source of truth", and that the host's live map belongs to the header only.streamStatusis already the last snapshot at-or-before the position, soif (isCleared && isReplay) → undefined, else paneStatuskeeps one source and fixes the bug. If you thinkliveTaskStatusis needed (for example for configure-time diagnostics that never reach the stream), explain why in the PR and update that comment.SourcePanel.tsx:365-385The DEAD ZONE comments ("pane data renders only when the position sits INSIDE a running track", "the snapshot panes gate on the same fact") become false with this change. Update them in the same PR. They also list tokens and flow as snapshot panes, which contradicts the PR body's claim that tokens is an accumulating pane. In the comment, say why errors leave the gate while tokens (:563) and flow (:523) keep it.- There's no test for the new logic. Once the selection is a pure function (see above), you don't need a harness for the hook. Add a unit test covering three cases: inside a track, a live dead zone (status is kept), and a replay dead zone (blank).
Should fix
SourcePanel.tsx:535-543Cut the comment to the why, in two or three lines. Drop the narrative of the old bug and theuseTaskEvents.ts:646line reference, which will go stale. The history belongs in the commit message.
| // status. Replay still blanks, because there the question genuinely is "what | ||
| // was true at this position". The accumulating panes below stay on `cleared`, | ||
| // where that gate is correct. | ||
| const snapshotStatus = cleared ? (isReplay ? undefined : liveTaskStatus) : paneStatus; |
There was a problem hiding this comment.
[blocking] Replace the nested ternary with if blocks (code-taste: readability), ideally inside an exported pure helper that you can unit-test. Also read from paneStatus here, not liveTaskStatus. The comment at :370-376 keeps the snapshot panes on the stream as the single source of truth, and streamStatus already holds the last snapshot at-or-before the position. Blank only when isCleared && isReplay; otherwise use paneStatus.
| // Errors/Warnings are a SNAPSHOT of the task's problems, not an accumulating | ||
| // window, so they must not vanish the moment a run ends. `cleared` is | ||
| // `!track.active`, and `active` means "inside a chapter that has not ended" | ||
| // (useTaskEvents.ts:646) — so gating these on it emptied both lists as soon | ||
| // as endTime was set, and the pane reported "No errors or warnings" about | ||
| // arrays it had just discarded. Outside a track, fall back to the task's own | ||
| // status. Replay still blanks, because there the question genuinely is "what | ||
| // was true at this position". The accumulating panes below stay on `cleared`, | ||
| // where that gate is correct. |
There was a problem hiding this comment.
[should fix] Cut this to the why, in two or three lines: errors/warnings are a status snapshot, so outside a live track they keep the latest one, and replay still blanks. Drop the bug narrative and the useTaskEvents.ts:646 reference, which will go stale. [blocking, related] Update the DEAD ZONE comments at :365-385, which still say every snapshot pane clears outside a track.
Summary
Type
fix
errorItemsandwarningItemswere gated oncleared, which is!track.active.activeis not "a run exists" —useTaskEvents.ts:646:It means "the position sits inside a chapter that has not ended". So as soon as a run finished and
endTimewas set,clearedwent true and both arrays were emptied one line before render:The pane then displayed "No errors or warnings" about arrays it had just discarded. A run that failed thirty seconds ago reported having raised nothing.
The strongest evidence that this is a mismatch rather than a deliberate choice is the pane's own empty-state text:
That is a correct description of a streaming accumulating window, and the wrong description for a status snapshot — which is what Errors and Warnings are. Updated in this PR to match the fixed behaviour.
Diagnostics raised when a node is configured rather than run have no chapter at all, so the same gate discarded them permanently. That case follows from the same mechanism.
Scope is deliberately narrow: gating on
clearedis correct for the accumulating panes and they are left alone — tokens (:562) andStatusPane(:629) still use it. Replay still blanks, because there the question genuinely is "what was true at this position".Testing
./builder testpassesNo test added, and I want to be straight about why rather than let the unchecked box pass silently:
SourcePanelis a 700-line component whose state comes fromuseTaskEvents, a live event-stream hook. Asserting this behaviour properly means driving that hook through a run's full lifecycle — open chapter, events,endTimeset — and there is no existing harness for it (packages/shared-ui's suite isscripts/*.test.mjsplus a handful of*.test.tsx, none of which mount this component). Building one is a larger piece of work than the three-line fix and I did not want to bundle it in. If you would like it, I am happy to do it as a follow-up — say the word.Also worth stating plainly: this was traced by reading the source, not by instrumenting a live session. The completed-run case follows directly from
active's definition and I am confident in it; the configure-time case is inferred from the same mechanism rather than separately observed.Checklist
Separate finding, deliberately not fixed here —
ProjectView.tsx:570passesliveTaskStatus={runKind === 'dev' ? statusMap[src.id] : undefined}, so the DEPLOY tab has no live status at all regardless of this gate. It is a different cause and belongs in its own change (and #1782 is already touching that file). Flagging it because it will look like this bug when tested.Linked Issue
Fixes #1832
Summary by CodeRabbit