Skip to content

fix(chat): closed-session notice names the link text and keeps the full stop - #20214

Merged
bolichen97 merged 1 commit into
mainfrom
fix/closed-session-notice-18588
Oct 14, 2026
Merged

bolichen97 merged 1 commit into
mainfrom
fix/closed-session-notice-18588

Conversation

@bolichen97

@bolichen97 bolichen97 commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: When a link to a closed session cannot be opened, the reader sees a notice that names the link they clicked, keeps the sentence whole, and tells them when the session was deleted.

Three defects in the inline notice a ?sid= link shows when the closed session's check fails:

  1. The notice says Couldn't open chat-7-1784661951. for a link whose text is the old run. The reader never saw that key.
  2. The notice is inserted between the link and the sentence's full stop, so the period lands after Try again and, at a phone width, wraps onto a line of its own.
  3. When Try again finds the session deleted (a 404), the link quietly turns muted and nothing is said.

Why it matters

A reader who clicks a link in an old transcript is told about a key they cannot match to anything, reads a sentence split by a red notice with its period orphaned, and after a retry that meets a deleted session gets no answer at all, so they cannot tell whether the retry worked. The same deleted session clicked as a chip already gets a "was deleted" banner; the retry path should say the same.

Not a goal

  • Changing the notice for a backticked session key: there the visible text IS the key, so it already names what the reader sees.
  • Moving punctuation around any other kind of link (external, forge, Jira, path): only the session link renders something after itself.
  • Changing what the probe fetches or when it runs.

What changed (motivation → approach → change)

Reader sees Before After
Notice name raw key probe title, else link text, else key
Period after Try again right after the link or chip
Notice inline, inside the sentence below the paragraph; the paragraph stays whole
Retry meets 404 silence, chip drops "was deleted" banner, then chip drops

useSessionChip takes an optional label (the link's visible text). A failed target's title is the probed title when a probe answered, else that label, else the key, and MdAnchor hands that title to ClosedSessionNotice instead of the key.

A new rehype pass, rehypeSessionRefPunctuation in treeTransforms.ts, runs after the sanitizer. For a session reference (an anchor whose href names a chat session, or a code span whose whole text is a session key) it moves the sentence punctuation at the start of the next text node onto the element as data-session-trailing. MdAnchor and InlineCode render that text right after the link or chip and before the notice, in every branch they can return; InlineCode reads it before its data-session* filter, so it never reaches the DOM. An authored data-session-trailing in raw HTML is stripped first, the way data-fenced is, so a message cannot make a reference print text it does not end with.

ClosedSessionNotice renders below the paragraph. A paragraph that holds a session reference (holdsSessionRef) draws an empty div[data-paragraph-notices] right after itself and provides it through ParagraphNoticeSlotCtx; the notice portals into it, so the paragraph's text is never interrupted, whether the reference ends a sentence or sits in its middle. Where no paragraph provides a slot (a tight list item, a heading, a table cell) the notice renders in place as a line of its own, after the reference's trailing punctuation. Other paragraphs keep their exact DOM; the golden DOM snapshot pins the slot on the two fixtures that hold a session link.

A definite "gone" is shared: when one reference's click or retry finds the session deleted, every mounted chip or link for that key drops together, notice included, so no second notice keeps offering Try again for a session the banner just called deleted.

In the retry, a fresh lookup that answers "no such session" goes through ClosedSessionCtx.open with the display title, the same call a click on a deleted chip makes. The chat page's open shows the "was deleted" banner and answers gone, which drops the target as before. useSessionChip is shared, so a backticked session chip's Try again takes the same hand-off, named by its key (the text the chip shows).

textOf in markdown/elements.tsx is exported so the anchor can read its label. website/capture/closed-session-notice.* is the isolated capture entry and shoot-closed-session-notice.mjs the script that produced the screenshots below.

Backwards compatibility

Compatible: every existing call of useSessionChip passes two arguments and gets the same answers; the rehype pass only touches text that directly follows a session-shaped ?sid= href, and the moved punctuation is rendered back in place by the anchor, so rendered text is byte-identical when no notice is up.

Removes nothing: the diff adds a parameter, a plugin and a prop; it deletes no path, key, record or public function.

Tests

website/src/test/MarkdownRenderer.closedSessionChip.test.tsx:

  • a prose-labelled ?sid= link's notice names the link text (Couldn't open the old run.), not the key;
  • the notice prefers the probed title when the background probe answered and the click-time check failed;
  • a ?sid= link with no prose label falls back to the key;
  • the full stop after a ?sid= link stays the link's own next sibling, the notice follows it, and the rest of the sentence follows the notice;
  • a link with no trailing punctuation keeps the notice right beside it;
  • punctuation after an ordinary (non-session) link is left where it was;
  • an authored data-session-trailing in raw HTML is dropped, not printed;
  • a retry that finds the session deleted calls ClosedSessionCtx.open with { key, title: 'the old run' } and the chip drops;
  • a backticked chip's retry on a deleted session takes the same hand-off, named by its key;
  • the full stop after a backticked session key stays the chip's next sibling; the notice renders below the paragraph;
  • a link in mid-sentence leaves the sentence whole; the notice renders below the paragraph;
  • a paragraph without a session reference draws no notice slot;
  • without a paragraph slot (a tight list item) the notice renders in place, after the trailing punctuation;
  • punctuation after an ordinary code span is left where it was;
  • a retry that finds the session deleted drops every reference to it, notices included;
  • the notice is block-level (flex, not inline-flex).

MarkdownRenderer.domGolden.test.tsx: the two fixtures holding a session link now pin the empty notice slot after their paragraph.

Red on the base for the three defects (4 of the new cases fail on origin/main), green with the fix: 32/32 in the file. The 60 MarkdownRenderer / InlineCode / ChatPage.closedSessionProvider test files (950 tests) pass; tsc -p tsconfig.app.json and eslint are clean.

Manual verification

Captured with node capture/shoot-closed-session-notice.mjs against this branch and against origin/main with the same harness: click the link, click the backticked key, then Try again, at 820px and 420px, dark and light.

Screenshots / video

Before, 420px: the notice quotes the key and the period sits on its own line.

before, narrow

After, 420px: the period stays with the link; the notice names the link text and follows the sentence.

after, narrow

After, 820px: the notice sits below the paragraph; the sentence reads whole.

after, wide

After, 820px, the same session as a backticked key: its period stays with the chip, its notice below its paragraph.

after, wide, link and backticked chip

After, Try again on a deleted session: the page's "was deleted" banner, then the chip drops.

after, retry on a deleted session

More

Before, 820px:

before, wide

Before, Try again on a deleted session: nothing is said, the chip just drops.

before, retry on a deleted session

After, light theme, 820px:

after, light

Related Issues

Closes #18588

Pattern harvest

Rule candidate: review-prompt
Pattern: an inline notice rendered as a React sibling of an anchor or chip splits the sentence; render it below the paragraph (portal to a slot the paragraph owns), with in-place-after-punctuation as the fallback.

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

@bolichen97
bolichen97 requested review from a team and KiroCrewCaptain as code owners October 14, 2026 08:43
@bolichen97
bolichen97 requested a review from mbriones98 October 14, 2026 08:43
@bolichen97
bolichen97 enabled auto-merge (squash) October 14, 2026 08:43
@bolichen97

Copy link
Copy Markdown
Contributor Author

CI triage on head 3eed4ec:

[operator: bolichen97]

@github-actions github-actions Bot added the readiness: action required A blocking check or review needs attention label Oct 14, 2026
@github-actions

github-actions Bot commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of 10a8182b359bc4d585db3a7ad49e7bcc22372281 — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

First-Principles-Verdict: PASS

Check one thing: the punctuation pass is now needed only where there is no paragraph slot (tight list items, headings, table cells). Keep it only if those places need it.

What this change ships

Inventory (10 items) — 10 justified

Intent: when a reader clicks a link to a closed session and it can't be opened, the notice should name what they clicked, leave the sentence whole and say when the session was deleted. This is a FIX (Closes #18588).

  1. The notice names the session by its probed title, else the link text, else the raw key. — justified
  2. The notice moved from inside the sentence to a line below the paragraph. — justified
  3. A full stop or comma after a session link or backticked key stays next to it. — justified
  4. With no paragraph slot (tight list item, heading, table cell), the notice is a line of its own after the punctuation. — justified
  5. When Try again finds the session deleted, the page now shows its "was deleted" banner. — justified
  6. When one reference finds the session gone, every reference to that key drops, notices included. — justified
  7. A data-session-trailing attribute written in raw HTML is removed before rendering. — justified
  8. Each paragraph that holds a session reference now has an empty notice slot after it. — justified
  9. textOf is now exported so the link can read its own label. — justified
  10. A capture page and a screenshot script for this notice. — justified

[FIRST-PRINCIPLES-REVIEWED] 10a8182

Description read

[DESCRIPTION-READ] a89d9d926e231acdbd9089bd7875572ceecfa35c893895df46d833db9aef5679

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 10a8182b359bc4d585db3a7ad49e7bcc22372281 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Nothing to check.

Suggestions

  • Most of the punctuation-moving pass is only needed where there is no slot. Inside a paragraph, the portal already puts the notice below the text, so the full stop has nothing to split it. If you give tight list items, headings and table cells the same notice slot, you could probably drop both rehypeSessionRefPunctuation and the keep(...) wrapping on every branch of MdAnchor and InlineCode.

[DESIGN-REVIEWED] 10a8182

Description read

[DESCRIPTION-READ] 4ad7d3d3789b76db36b7639c310827c614750540f4c95da39a828c18c58f4834

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 10a8182b359bc4d585db3a7ad49e7bcc22372281 — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 10a8182

Verdict parsed from the review's SHA-scoped output markers for commit 10a8182b359bc4d585db3a7ad49e7bcc22372281.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 10a8182b359bc4d585db3a7ad49e7bcc22372281: <one-sentence reason>

@github-actions

github-actions Bot commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ✅ PASS

UX-level review of 10a8182b359bc4d585db3a7ad49e7bcc22372281 — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: PASS

Check shot-04: one session raises two notices with two different names ("the old run" and "chat-7-1784661951"), and the blind reader couldn't tell whether that was one failure or two.

[UX-REVIEWED] 10a8182

Description read

[DESCRIPTION-READ] c1c12e2f50ef856797822645f9e16a7b5926eec44e523bc60f194446403358dc

That sha256 names everything this verdict read: the PR description, plus the 11 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 10a8182b359bc4d585db3a7ad49e7bcc22372281 and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 10a8182

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 10a8182b359bc4d585db3a7ad49e7bcc22372281: <one-sentence reason>

@bolichen97
bolichen97 force-pushed the fix/closed-session-notice-18588 branch from 3eed4ec to 9e7a2e7 Compare October 14, 2026 10:18
@bolichen97

bolichen97 commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor Author
  • span=d618d7cfdc7a The notice still splits the paragraph at wide widths — fixed in 9e7a2e7

ClosedSessionNotice now carries mx-1.5 instead of ml-1.5, so "Try again" keeps a gap before the next word of the sentence; the re-shot wide screenshot in the description shows it.
Pinned by the new test the notice keeps a gap on both sides, so Try again does not run into the next word.

[operator: bolichen97]

@bolichen97

bolichen97 commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor Author
  • span=77ab287db3c9 The inline notice and the banner read as two different failures — rebutted

They are two different outcomes, in order: "Couldn't open the old run." is a check that FAILED (gateway unreachable), with Try again as the way forward; the banner "was deleted and cannot be opened" is the retry's definite answer that the session is gone, after which there is nothing to retry.
The banner is the page's existing session_gone_open_failed_named string, the same one a click on a deleted chip shows; #18588 item 3 asks for exactly that routing. Re-wording either message is a separate copy change outside this fix.

[operator: bolichen97]

@bolichen97

bolichen97 commented Oct 14, 2026 •

Copy link
Copy Markdown
Contributor Author
  • span=bfcd31a21060 Item 4 — undeclared: the backticked chip also gets the retry-404 hand-off — fixed in 9e7a2e7

Declared in the description ("useSessionChip is shared, so a backticked session chip's Try again takes the same hand-off, named by its key") and pinned by the new test a backticked chip whose retry finds the session deleted takes the same deleted hand-off, named by its key.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • Does a retry that hits a deleted session still drop the chip when open answers something other than gone? — rebutted (confirmed by reading openRow)

openRow handles every open outcome: gone caches null, clears the notice and settles null (chip drops); failed settles failed and shows the notice with Try again (the check failed, so the chip must stay for another retry, not drop); opened clears the notice and the resumed session enters the roster, where the live path wins on the next render. The removed settle(null) only ever fired before open had judged the key, which is what left a retry answered by nothing.
Pinned by a retry that finds the session deleted goes through the page's deleted banner (gone), a retry that fails again keeps the notice up (failed) and a click on a chip whose probe failed shows the inline notice, and retry opens it (opened).

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author

Rebased: main moved in website/src/components/MarkdownRenderer.tsx and website/src/components/ since this head went green (30bb0b3 -> 641b42a, carrying the #20212 fix for the ChatEmbed.optimisticSend red); CI re-running on 9e7a2e7. [operator: bolichen97]

@bolichen97
bolichen97 force-pushed the fix/closed-session-notice-18588 branch from 9e7a2e7 to 5fa9786 Compare October 14, 2026 10:42
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 14, 2026
@bolichen97

Copy link
Copy Markdown
Contributor Author
  • The notice still splits the paragraph — fixed in 5fa9786

ClosedSessionNotice is now block-level (flex, a span so it stays legal inside the <p>): it takes a line of its own below the sentence that named the session, after that sentence's full stop, so the paragraph reads sentence / notice / next sentence, at every width.
Pinned by the notice is a line of its own, not an inline run between two sentences; the re-shot wide screenshots in the description show it.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • The deleted link still looks live — rebutted

After gone the link IS in the not-live style: useSessionChip returns null, so MdAnchor renders it text-muted with no underline, the same muted treatment every ?sid= link to a session the reader cannot open already gets (pinned by a retry that finds the session deleted goes through the page's deleted banner, which asserts text-muted). The muted-vs-live palette of session links is the existing design for a not-open link, not something this PR introduces; a line-through or plain-text rendering for all not-open links is a separate design change.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • The punctuation pass only handles <a> session links; a backticked chip renders the notice before its full stop — fixed in 5fa9786

The pass is now rehypeSessionRefPunctuation: it also moves the punctuation after a code span whose whole text is a session key, and InlineCode renders it between the chip and its notice (and in every other branch it can return), reading the marker before its data-session* filter so it never reaches the DOM.
Pinned by the full stop after a backticked session key stays with the chip; the notice follows it and punctuation after an ordinary code span is left where it was; the new link-plus-chip screenshot in the description shows both.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • On a retry 404, does the page's open only re-check and show the banner, never start or resume a session? — rebutted (confirmed by reading ChatPage.closedSessionActions.open)

open never creates a session. It re-reads api.sessionMeta(key): null answers the banner and gone (the retry's case, pinned by a retry that finds the session deleted goes through the page's deleted banner and the page test a click on a session deleted since says it was deleted); a row answers resumeFromHistory on an EXISTING transcript, which is the right outcome when the session came back between the two reads, since the reader pressed Try again to open it; a thrown read answers failed and the inline notice stays up.

[operator: bolichen97]

@bolichen97
bolichen97 force-pushed the fix/closed-session-notice-18588 branch from 5fa9786 to 0c57b38 Compare October 14, 2026 11:05
@bolichen97

Copy link
Copy Markdown
Contributor Author
  • The block-level notice appears right after the link, not after the sentence; a mid-sentence link splits the sentence across three lines — fixed in 0c57b38

The notice now renders BELOW the paragraph: a paragraph holding a session reference draws an empty slot div[data-paragraph-notices] after itself (ParagraphNoticeSlotCtx) and ClosedSessionNotice portals into it, so the paragraph's text is never interrupted, whichever position the link or chip has. In-place after the trailing punctuation remains only as the fallback where no paragraph exists (tight list item, heading, table cell).
Pinned by a link in mid-sentence leaves the sentence whole; the notice renders below the paragraph, a paragraph without a session reference draws no notice slot and the tight-list fallback test; the golden DOM pins the slot.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • The punctuation pass only covers session references; the two reveal-error notices split sentences the same way — rebutted (outside the frozen goal)

The reveal-error notices belong to the confirmed PATH chip and path link (useRevealFailure: a Shift+click reveal the gateway refused), a different surface with its own dismiss control and its own tests, unchanged by this diff. The frozen goal is the closed-session notice (#18588), and "Not a goal" excludes moving punctuation around any other kind of link. Giving the path-reveal notice the same below-paragraph slot is a one-line follow-up once this mechanism has landed, not a defect in code this PR adds.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • After Try again finds the session deleted, the banner says "was deleted" while another notice for the same session still offers Try again — fixed in 0c57b38

A definite "gone" is now shared: markGone records it and notifies every mounted useSessionChip, and a shared null wins over a chip's own failed, so every link and chip for that key drops together, notice included, the moment one of them learns the session is deleted.
Pinned by a retry that finds the session deleted drops every reference to it, notices included.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • span=ea53908d3b06 The punctuation pass duplicates the portal — rebutted

The two mechanisms serve two placements, not one. The slot exists only where a <p> wraps the text; a session reference in a TIGHT list item, a heading or a table cell has no paragraph, and there the notice still renders in place, where without the pass the period lands after Try again again (the original defect). Pinned by without a paragraph slot (a tight list item) the notice renders in place, after the trailing punctuation: remove the pass and that test goes red. Agent-written transcripts put session keys in tight bullet lists constantly, so the fallback is a real path, not a theoretical one.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • span=6fcffaaf40e6 Let the nearest block container (p, li, h, td) provide the notice slot* — rebutted (disproportional for this fix)

Slots on every block container means a stateful wrapper and an extra DOM element on four more element overrides (li, h1-h6, td, each with its own existing override in elements.tsx), plus the golden DOM for each, to move a notice that in those containers already renders whole after the sentence's punctuation. The paragraph is where transcript prose lives; the other containers keep the in-place fallback with the sentence intact. Widening to a generic block-slot mechanism is a separate change if a second notice ever needs it.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • span=5318b900608b "only the session link renders something after itself" is untrue: the two path reveal-error notices — rebutted (outside the frozen goal; the wording is granted)

Granted: the path chip and path link render an inline reveal-error notice after themselves too, so that clause of Not-a-goal overstates. The scope decision it records stands: the goal is #18588's closed-session notice, and the path reveal notice is a different surface (useRevealFailure, its own dismiss, its own tests) this diff does not touch. Giving it the same paragraph slot is a one-line follow-up once this mechanism has landed, not a defect in code this PR adds; the frozen sections are not edited by the agent.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • span=6823598912f8 Two names for one session: the link's notice says "the old run", the chip's says the key — rebutted

Each notice names the thing the reader clicked, in the words they saw: that is #18588 item 1 ("Name what the reader sees"). A link shows prose, so its notice says the prose; a backticked chip shows the key, so its notice says the key. Renaming the chip's notice to a title it does not display would make the notice disagree with the chip beside it. Two references to one session in one bubble is the author's text, and each is answered in its own words.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • span=f39181c69c25 The full stop after the chip sits apart from it — rebutted (pre-existing chip geometry)

The gap is the chip's own horizontal padding (px-1.5 on every inline-code chip), present on main for any chip followed by punctuation (a path chip, a copy chip) and unchanged by this diff. The period is the chip's immediate next sibling in the DOM (pinned by the full stop after a backticked session key stays with the chip); it no longer lands after the notice, which is what the issue asked. Changing the chip padding affects every code span in every transcript and is outside this fix.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • span=989c32270992 Two Try again buttons; nothing says they act on the same thing — rebutted (mild, new scope)

Each notice belongs to the reference under which it sits, and pressing either one resolves both (a gone or an opened answer drops every reference to that key, pinned by a retry that finds the session deleted drops every reference to it, notices included). Collapsing identical notices across references is a new de-duplication feature, not part of #18588's three items, and a reader who presses the second button after the first gets the same correct outcome.

[operator: bolichen97]

@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 14, 2026
@bolichen97
bolichen97 force-pushed the fix/closed-session-notice-18588 branch from 0c57b38 to 10a8182 Compare October 14, 2026 11:32
@bolichen97

Copy link
Copy Markdown
Contributor Author

Rebased: main moved in website/capture/ and website/src/components/ (InfoTip.tsx) since this head went green (641b42a -> d06b611); related tests re-run locally (closedSessionChip, domGolden, InfoTip: 75/75), no code change; CI re-running on 10a8182. [operator: bolichen97]

@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: passed Eligible automated validation passed for the current revision labels Oct 14, 2026
@bolichen97

Copy link
Copy Markdown
Contributor Author
  • span=7cf5b35f0b94 Give tight list items, headings and table cells the same slot and drop the punctuation pass — rebutted (disproportional here)

Those three containers each have their own element override; a stateful slot wrapper plus an extra DOM node on each, and three more golden-DOM pins, to replace a pass that already keeps the sentence whole there. The pass is kept BECAUSE those places need it: without a paragraph slot (a tight list item) the notice renders in place, after the trailing punctuation goes red without it, and agent transcripts list session keys in tight bullets routinely. A generic block-slot is a reasonable follow-up once a second notice wants it.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • Is the punctuation pass still needed where there is no paragraph slot? — confirmed, kept for that reason

Yes: tight list items (- see \chat-7-…`.), headings and table cells have no

, so the notice renders in place there, and without the pass the period lands after Try again again. Pinned by without a paragraph slot (a tight list item) the notice renders in place, after the trailing punctuation`; inside a paragraph the pass is inert for layout (the notice is below) and only keeps the DOM order identical in both placements.

[operator: bolichen97]

@bolichen97

Copy link
Copy Markdown
Contributor Author
  • Shot-04: one session raises two notices with two names — rebutted

Each notice names what the reader clicked, in the words shown there (#18588 item 1: name what the reader sees): the link's prose for the link, the key for the backticked chip, since the chip displays the key. Both buttons act on the same session, and either Try again resolves both notices (a gone or opened answer drops every reference to that key, pinned by a retry that finds the session deleted drops every reference to it, notices included). Renaming the chip's notice to text the chip does not show would make the notice disagree with the chip beside it.

[operator: bolichen97]

@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 14, 2026

@KiroCrewCaptain KiroCrewCaptain left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Tier 1 auto-approve: fix (12 files). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, security checklist all-NO, no reviewer-lane findings, readiness signal fresh for this head SHA, declared category matches the diff. Category: the closed-session notice names the link text and renders below the paragraph, so the sentence keeps its full stop; covered by new MarkdownRenderer tests.

@bolichen97
bolichen97 merged commit e4270cf into main Oct 14, 2026
102 checks passed
@bolichen97
bolichen97 deleted the fix/closed-session-notice-18588 branch October 14, 2026 13:08
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 14, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

fix(chat): closed-session notice names the link text and keeps the full stop

2 participants