You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
console: System Hub 计数失败后「无徽章」与「这张卡本来就没有计数」「还在加载」三态同形 —— 403 的管理员得到的是沉默,不是「你没有权限」 #3685
Findings triage — verdict: HOLD (finding retained).
Why still held, and explicitly not because the severity is low: the card is right that the outcome is missing information at the moment it is most needed — an admin without sys_audit_log read access now sees an Audit Log card pixel-identical to Settings. Nothing here is being graded down. What blocks promotion is that the fix is a fork whose answer is already owned by an open decision: #3655's option C retires this bespoke card wall entirely (SystemHubPage's docblock already carries @deprecated), and the body correctly notes this card is void if C is chosen and stands if A or B is. Promoting now would put a second decision in front of the maintainer that is downstream of one they already hold, and would risk spending an implementation cycle on a page slated for removal.
Restart condition — the #3655 ruling, and it resolves this card in both directions:
Verified independently of the body:#3679's contribution stands as described — the badge renders under {card.count !== null && …}, so falling back to null restores an existing degraded form rather than inventing UI, which is why補 error copy was correctly out of that dispatch's authorised surface. The tri-state collapse (eight hardcoded-null cards / five counted cards pre-fetch / post-#3679 failures) is a property of that same conditional, not a regression #3679 introduced. The role="status"Loading statistics... at the page top does separate loading in time but not on the card, exactly as stated.
No target:<major>: the page is on the @deprecated path with its retirement under active decision; blocking an RC on a card wall that may not ship is not defensible. Revisit if #3655 lands on A/B.
Closed, not planned — absorbed by the #3743 ruling (2026-08-09).
The maintainer ruled option C on #3743: the bespoke SystemHubPage card wall retires in favour of a navigation/registry-driven entry (「全部接受」, decision-inbox round of 2026-08-09). This card's own boundary note anticipated exactly this: 「若维护者最终选 C,本单随之作废」. The three-states-one-pixel ambiguity dies with the wall; no error-copy work lands on a surface being deleted.
If the retirement implementation somehow preserves per-card counts on the new surface, the 403-silence concern re-files against that surface with fresh evidence — this card stays closed either way.
Activity
claude commentedon Aug 7, 2026
Findings triage — verdict: HOLD (
findingretained).Why still held, and explicitly not because the severity is low: the card is right that the outcome is missing information at the moment it is most needed — an admin without
sys_audit_logread access now sees an Audit Log card pixel-identical to Settings. Nothing here is being graded down. What blocks promotion is that the fix is a fork whose answer is already owned by an open decision: #3655's option C retires this bespoke card wall entirely (SystemHubPage's docblock already carries@deprecated), and the body correctly notes this card is void if C is chosen and stands if A or B is. Promoting now would put a second decision in front of the maintainer that is downstream of one they already hold, and would risk spending an implementation cycle on a page slated for removal.Restart condition — the #3655 ruling, and it resolves this card in both directions:
system的记录页 #3655 → A or B (the card wall survives): promote topm:queuescoped to the body's shape 1 — a distinguishable badge (variant="outline"—) withtitle/aria-labelcarrying the reason. Smallest surface, stays inside the existing Badge form, preserves the per-card isolation console: System Hub 计数的.catch(() => ({ data: [] }))把 500 / 401 / 403 / 断网都渲染成一个确定的0(404 根本到不了它) #3679 deliberately kept. Shape 3 (a page-level "some statistics failed to load") is on record as considered and rejected for flattening that isolation back out; shape 2 (per-card retry) is a widening that should be its own card if wanted.system的记录页 #3655 → C (card wall retires): close this as superseded, no work.system的记录页 #3655 stalls and a real 403-admin report arrives — a user hitting the silence outranks the tidiness of waiting.Verified independently of the body: #3679's contribution stands as described — the badge renders under
{card.count !== null && …}, so falling back tonullrestores an existing degraded form rather than inventing UI, which is why補 error copy was correctly out of that dispatch's authorised surface. The tri-state collapse (eight hardcoded-nullcards / five counted cards pre-fetch / post-#3679 failures) is a property of that same conditional, not a regression #3679 introduced. Therole="status"Loading statistics...at the page top does separate loading in time but not on the card, exactly as stated.No
target:<major>: the page is on the@deprecatedpath with its retirement under active decision; blocking an RC on a card wall that may not ship is not defensible. Revisit if #3655 lands on A/B.本评论来自分诊座位 Routine(#5474 试点),不构成认领。
Generated by Claude Code
os-project-manager commentedon Aug 9, 2026
Closed, not planned — absorbed by the #3743 ruling (2026-08-09).
The maintainer ruled option C on #3743: the bespoke
SystemHubPagecard wall retires in favour of a navigation/registry-driven entry (「全部接受」, decision-inbox round of 2026-08-09). This card's own boundary note anticipated exactly this: 「若维护者最终选 C,本单随之作废」. The three-states-one-pixel ambiguity dies with the wall; no error-copy work lands on a surface being deleted.If the retirement implementation somehow preserves per-card counts on the new surface, the 403-silence concern re-files against that surface with fresh evidence — this card stays closed either way.
Generated by Claude Code