Repository navigation
🐛 fix: show verified external merges at the terminal tracker step - #16036
Conversation
Signed-off-by: Andy Anderson <andy@clubanderson.com>
…rminal Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Source completeness follow-up f784e41: when raw nested_under points to an issue outside the current payload and no DOM host exists, terminal observation now falls back to the actual retained root PR row instead of losing the model. Added a focused no-host/no-DOM regression. Prior43d checks had no substantive failures, but they are not validation of f784; full CI is being observed again on the current exact head. No manual merge/deployment. |
There was a problem hiding this comment.
Correctness review (read at head f784e41, the current head; the dispatch cited 43d6b1f).
Pinned terminal rows bypass the active filter and page — tracker-merged.js:1040-1045, tracker-filter-render.js:174
Before this PR a pinned row could only be a node already on the board (stageTrackerPinnedRows fell back to el.querySelector([data-ham-key]) and dropped the pin when nothing was drawn), so it had necessarily passed stageTrackerVisible and pagination. Now stageTrackerTerminalObserve (tracker-merged.js:795-806) creates a pin for every fresh key whose row exists in _stageTrackerTerminalRows — i.e. any item in the last payload, filtered or not — and stageTrackerPinnedRows renders it from the model via renderItem with no filter check. renderStageTracker computes visible = stageTrackerVisible(d.items, f, ctx) (tracker-filter-render.js:92) but passes it => stageTrackerRowHtml(it, stages, ctx) unconditionally (:174), so with f.repo = hivecommons/console a hivecommons/hive merge appends a hive row to the bottom of the console-filtered list for STAGE_TRACKER_CELEBRATE_PIN_MS; same for the kind tab (a PR row under "issues"), search, labels, and rows from other pages. This contradicts the existing contract in stageTrackerCelebrateMerges ("filtered out, other page … its row is celebrated before removal only when it is drawn", tracker-merged.js:981-984) and stage_tracker_external_terminal_test.go has no filter case.
Suggested fix — gate the model render on the filter, keeping the pin (so stageTrackerTerminalEvents and the release still run) but not drawing it:
--- a/src/pkg/dashboard/static-src/js/tracker-filter-render.js
+++ b/src/pkg/dashboard/static-src/js/tracker-filter-render.js
@@
- const rows = stageTrackerPinnedRows(el, [{ key: STAGE_TRACKER_HEADER_KEY, html: stageTrackerHeaderHtml(stages) }].concat(inlineRows), nowMs, it => stageTrackerRowHtml(it, stages, ctx));
+ // A pinned terminal model is drawn only where its live row would have been: same filter, never another page's row.
+ const rows = stageTrackerPinnedRows(el, [{ key: STAGE_TRACKER_HEADER_KEY, html: stageTrackerHeaderHtml(stages) }].concat(inlineRows), nowMs,
+ it => stageTrackerMatches(it, f, ctx) ? stageTrackerRowHtml(it, stages, ctx) : '');
--- a/src/pkg/dashboard/static-src/js/tracker-merged.js
+++ b/src/pkg/dashboard/static-src/js/tracker-merged.js
@@ function stageTrackerPinnedRows(el, rows, nowMs, renderItem) {
const item = c.kind === 'merged' && typeof renderItem === 'function' && typeof stageTrackerTerminalPinnedItem === 'function' ? stageTrackerTerminalPinnedItem(k, nowMs) : null;
if (item) {
- out.push({ key: k, html: renderItem(item), pinned: true });
+ const html = renderItem(item);
+ if (html) out.push({ key: k, html, pinned: true });
stageTrackerScheduleRelease(c.until - nowMs);
continue;
}plus a regression case in stage_tracker_external_terminal_test.go where renderItem returns '' and asserts stageTrackerPinnedRows(...).length === 0 while _stageTrackerCelebrating[key] survives.
Advisory (not blocking): stageTrackerCarryOpen now keeps every true key for the life of the tab and in sessionStorage (tracker-merged.js:53-58, :94); growth is bounded by merges seen per session so it is fine in practice, but worth a comment since the prior comment promised "a few keys".
— hive: agent=adjudicator backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor. Review findings from Read at What I verified:
One open question, not a finding: once a PR's Note: this and #16037 both regenerate |
|
FINAL immutable exact-head CI receipt: f784e41 has all17 applicable workflow runs terminal SUCCESS/SKIPPED, all79 CheckRun contexts success/skipped, DCO success. v2 Tests41202574593, v2 CI41202574552 (includes Node), security41202574466, image41202574666 including ARM freshness, dashboard lint41202574527 all SUCCESS. Current-head Hive reviewer-2 independently reports no concrete defect and verifies the action seam, per-PR actor identity, irreversible carry and closed-linked terminal exception. PR remains OPEN under Hive-managed review/preparation; no manual merge or deployment by me. Existing short pin expiry intentionally removes stale-open representations rather than regressing a verified merged PR. |
|
Thank you for your contribution! Your PR has been merged. We'd love to hear how your experience was: share feedback |
|
MERGE RECEIPT: explicitly authorized guarded admin squash matched the tested f784e41 and merged source cf008a5 by clubanderson at2026-10-31T13:14:26Z. Full17 workflows/all79 checks terminal pass/skip before merge; no actual human hold, no gate weakening. DCO trailer present in squash. No deployment/runtime activation by me. |
What changes
Verified canonical merge facts now project root, inline and nested PRs directly to the existing terminal110 step, even when the browser never observed queue admission. Retained models survive bounded ledger lag; actual merger identity is preserved and unknown remains unknown. Related issues are not implicitly closed. No synthetic100, native queue membership, or Hive/manual-click provenance.
Correctness
Per-PR immutable observation/windows prevent one merged sibling renewing another. Existing release timers rearm for already-rendered hosts. Closed terminal PRs retain navigation but stale captured write actions are rejected at the final request seam. History-only unresolved siblings never reappear as visible work.
Review and validation
Independent code review found stale actions, sibling windows, release scheduling, duplicate fixture constant and unresolved-history resurrection; all were fixed with focused regression cases. Only changed-file gofmt and required dashboard generation ran locally, per AGENTS. CI is the sole verdict; exact-head terminal result will be recorded here. No deployment, live policy/config changes or manual merge requested.