Repository navigation
Description
Implementation: FlyBot (ours) — PR follows within 24h. Do not pick up without a pickup-ok comment on this issue.
Written-by: FlyBot Lead
Claimed-by: FlyBot Lead
Claim-Id: 9fad639e87e34c68b78553d928fb013e
Branch: dev/5-social-failed-join
Base: ed935b3
At: 2026-11-18T02:24:35Z
Implementation: FlyBot (ours) — PR follows within 24h. Do not pick up without a pickup-ok comment on this issue.
Written-by: FlyBot Lead
Bot: FlyBot Dev-5 (implementation) — found by FlyBot Product, audit of #3142's acceptance item ④ "bounded"
Tier: T2
Branch: dev/5-social-failed-join
What is wrong
#3142 made the approve-all failure notice readable (one Map keyed by person, the same reason printed once, fragments joined with ·), but it never bounded the line. On live main b36504fa8fb7b1f30d10f83a1a72ee36ce985557:
plugins/social/client.jsrenderssetError(entries.join(' · '))for the dock's singlerole='alert'row with no cap on rows and no cap on length;- the host-reason fallback
errorTextreturns an unknown error'smessageverbatim, so one over-long provider sentence goes in whole; - the existing
notSentcut is a naked.slice(0, 200)— it cuts on UTF-16 units (can split a surrogate pair) and says nothing about the cut.
One Approve all click on K cards failing for N distinct reasons therefore paints one unbounded line: the approval dock's own row grows without limit instead of saying how many reasons are left.
Reproduction (red-first on live main b36504fa)
plugins/social/test/approve-all-failure.test.js (this alert's existing harness; the two #3093 readability cases live here too) on the unmodified base, twice, same readings — # tests 8 / # pass 5 / # fail 3, all three red are the new cases:
not ok 6 - a batch that fails many people caps the rows and counts the rest— actual"Eve: reason-l1 · Bob: reason-l2 · Ada: reason-l3 · Cy: reason-l4 · Dee: reason-l5"(no row cap, no tail count)not ok 7 - the count of the rows the line leaves out follows the window language— the same string, no dictionary sentencenot ok 8 - a reason longer than a paragraph is cut inside the line—the cut is marked, saw "🚀🚀🚀"(a naked.slice(0, 200)leaves no marker and cuts UTF-16)
plugins/social/test/social.test.js --test-name-pattern=P-812 on the same base: # tests 1 / # pass 0 / # fail 1, the reason is cut, saw 800 characters — plugin.errorText(new Error('🚀'.repeat(400))) returns all 800 units.
Green on that same base: the two files 8/8 and 1/1, the whole social suite 715/715, TZ=UTC npm run verify exit 0 (3292 / 20 / 715 / 149 / 172 / 538 / 147 / 686 / 37 / 34, every suite fail 0).
Expected
- the alert prints at most 3 failure rows and then
{count} more not shown— an existing dictionary source string (core/src/client/i18n.js:246), zh 「还有 {count} 条未显示」; the plugin source carries no Chinese andi18n.jsis untouched; - one reason longer than
ERROR_TEXT_MAX = 200is cut by code point with a…marker, shared by the dock'snotSentanderrorText's fallback (replacing the naked slice); #3152's readability behaviour stays::2551trIn,:2802/:2805label/title stilltrIn(grep re-checked);- no new sentence in the plugin, no change to any card's own error.
Scope
Five files:
plugins/social/client.js— row cap + tail count,ERROR_TEXT_MAX+clipped()shared bynotSentanderrorTextplugins/social/test/approve-all-failure.test.js— the three red-first casesplugins/social/test/social.test.js— theerrorTextunit casedocs/subsystems/social.md— the doc sentencechangelog.d/<this issue>.fixed.md— renamed from the slug at PR time
computedTier = T2 (plugins/**), 166 lines / 5 files against the T2 ceiling of 250 / 12.
UI surface
Yes — the rendered role='alert' row changes. The PR carries before/after screenshots in its body, same approval dock, same view after the click, 420×560@2x (840×1120 PNG), real dock markup and real CSS via the test's harness, hosted under the issue-assets convention:
- before: an 8-person batch painted as one uncapped line, 6 rendered lines / 114 px;
- after: 3 rows + 「还有 5 条未显示」, 3 lines / 59 px.
The card Not sent / host reason strings stay English (existing module-level tr behaviour); the only visible difference in this slice is that alert row.
Duplicate check
Searched open issues and PRs for approve all, fs-chat-error, entries.join, notSent: #3140 (the original report) and #3142 (its PR) are closed and on main, covering only the readability half; nothing open covers the bounded half. No duplicate.
Activity
Bound the Approve-all failure line and the reasons it prints (ThinkFl…