[ThomasK33]

Fixes #7637

Summary

geist 1.8.1 (from #7625) made Geist Mono's text box taller: ascent/descent went from 955/245 to 1005/295 units, so a 12 px code line's text box grew from 14 px to 16 px. The phone static workspace view paints the seeded code line before Geist Mono loads (fallback monospace, 14 px tall). The live app's code line now paints larger than that copy, so Chrome records a second, later LCP entry and Lighthouse simulates LCP at the end of the JS load. This PR adds ascent-override/descent-override/line-gap-override to the Geist Mono @font-face, which restores 1.5.1's vertical metrics and keeps the bump.

Background

• Static view (🤖 perf: the static workspace view's early script (not injected yet) #6879/🤖 perf: inject the static workspace view on a phone's workspace page #6880): on a phone /workspace/<id> the early script paints the transcript tail before the app loads. Geist Mono is deliberately not preloaded (🤖 perf: compress the SPA page and preload its first-view fonts in xum server (T3 C step 1) #6216, comment in staticAssets.ts), so code in the static view paints in the fallback monospace.
• Before 🤖 fix: bump effect to stable 4.0.4 and dependencies within range #7625, the live code line (Geist Mono 1.5.1, 178x14 px text box) was no larger than the static copy (fallback, 180x14), so the static paint stayed the final LCP. After 🤖 fix: bump effect to stable 4.0.4 and dependencies within range #7625 it is 175x16, which is larger.
• Font tables (fontTools, unitsPerEm 1000, fsSelection.USE_TYPO_METRICS set in both):

geist	hhea/typo ascent	descent	lineGap	content height
1.5.1	955	245	0	1.2 em
1.8.1	1005	295	0	1.3 em

The ascent minus descent is the same (710), so the overrides move no glyph: the baseline in a fixed line-height stays where it is. They only shrink the inline text box back to 1.2 em, which also restores the pre-bump height of inline-code backgrounds and of line-height: normal monospace text (for example the terminal default font).

Bisect

Check: seed like the auditor (seed.mjs, 4 text rows), run xum server with env -i, XUM_DISABLE_TELEMETRY=1, a temp XUM_ROOT and a free port, then open the workspace page in headless Chromium (Playwright; 412x823, DPR 1.75, isMobile, hasTouch, a Moto G Power UA, service workers blocked, a fresh browser per run). It records every largest-contentful-paint entry. PASS = no entry outside the static view after the static paint.

Build	Result	LCP entries (time, size, element)
6959dfc (base)	PASS 3/3	392 ms 9306 static code line (no later entry)
0c9a0c0 (main)	FAIL 3/3	352 ms 9306 static, then 728 ms 9555 live code line (runs 2 and 3: 60 → 416 ms, 56 → 388 ms)
0c9a0c0 + geist 1.5.1 only	PASS 3/3	400 ms 9306 static (no later entry)
this PR	PASS 5/5	348 ms 9306 static (no later entry); live text box 175x14

Reverting geist alone flips the check, so I did not bisect the Tailwind pair or the other packages.

Validation

• LCP-entry check above: head PASS 5/5, main FAIL 3/3.
• make first-run-view-parity: parity.json is identical on base and on this PR. The paragraph boxes match, and both sides show the same shell pixel diff in one 14x14 icon box, so that diff was there before this PR.
• make check-phone-chrome: the phone skeleton matches the live page.
• make static-check passes.
• Built CSS has the three descriptors in the Geist Mono @font-face.

No test added. The defect is a rendered text-box size in a real browser with real font files. A happy-dom or build-time check cannot measure that, and a test that reads the CSS would only repeat the three values. The browser check above is the reliable gate, and the Lighthouse VM covers it after merge.

Risks

Low. The change is CSS only and touches the Geist Mono face only. Glyph positions stay the same. Text boxes of monospace text go back to their pre-#7625 height. Geist (sans) also changed its metrics in 1.8.1 (920/220/100 → 1005/295/0). I did not change it: it is preloaded, so the static view and the live app already paint it with the same font.

---

Generated with xum • Model: anthropic:claude-opus-5-5 • Thinking: high

[ThomasK33]

@codex review

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-26T19:41:12.669140Z	e659512	Manual request
🔒 Security Review	✅ Completed 2026-10-26T19:41:41.623249Z	e659512	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Can't wait for the next one!

Reviewed commit: e65951282d

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

🛡️ Codex Security Review · Automatically triggered

Security review completed. No security issues were found in this pull request.

Reviewed commit: e65951282d

View security finding report

Only the user who started this review can view the report in Codex.

ℹ️ About Codex security reviews in GitHub

This is an experimental Codex feature. Security reviews are triggered when:

• You comment "@codex security review"
• A regular code review gets triggered (for example, "@codex review" or when a PR is opened), and you’re opted in so security review runs alongside code review

Once complete, Codex will leave suggestions, or a comment if no findings are found.

[github-actions]

🟢 e2e stories (Haiku): 14 passed

22 agent steps · 7 replayed from cache · 15 model calls · 46.3k tokens (24% cached)
Spend: $0.09 of the $1.00 cap (proxy ledger)

All 14 tests in 9 files

	Test	Agent	Time
🟢	chatRestart.e2e.ts · 1 passed	1 step · 1 call	5.3s
🟢	a chat survives a restart of the server	1 step · 1 call	5.3s
🟢	compaction.e2e.ts · 1 passed	1 step · 1 call	4.3s
🟢	/compact replaces the old chat with a summary	1 step · 1 call	4.3s
🟢	pluginTrustRestart.e2e.ts · 1 passed	4 steps · 2 calls	9.5s
🟢	a trusted plugin hook stays trusted after the server restarts	4 steps · 2 calls	9.5s
🟢	settingsRoundTrip.e2e.ts · 1 passed	2 steps · 1 call	8.8s
🟢	a chat survives trips through Settings	2 steps · 1 call	8.8s
🟢	sidebarPerWorkspace.e2e.ts · 1 passed	4 steps · 1 call	4.2s
🟢	a new workspace starts with its right sidebar collapsed while another keeps it open	4 steps · 1 call	4.2s
🟢	smoke.e2e.ts · 1 passed	2 steps · 1 call	3.1s
🟢	the agent opens the playground workspace	2 steps · 1 call	3.1s
🟢	streamErrors.e2e.ts · 1 passed	1 step · 1 call	6.6s
🟢	stream errors show in the chat, and the chat recovers	1 step · 1 call	6.6s
🟢	todoNudge.e2e.ts · 1 passed	1 step · 1 call	3.6s
🟢	the agent finishes open TODO items after the turn-end nudge	1 step · 1 call	3.6s
🟢	touchAlignment.e2e.ts · 6 passed	6 steps · 6 calls	21.9s
🟢	the Settings close button is level with the title on a touch screen 390px wide	1 step · 1 call	4.3s
🟢	the Settings close button is level with the title on a touch screen 768px wide	1 step · 1 call	4.0s
🟢	Backup settings checkboxes are small and level with their labels on a touch phone	1 step · 1 call	4.5s
🟢	the heartbeat checkbox is small and level with its label on a touch phone	1 step · 1 call	3.3s
🟢	a reply's model and time stay on one line on a touch phone	1 step · 1 call	2.9s
🟢	the workspace actions menu stays on the title row on a touch phone	1 step · 1 call	2.8s

_{e2e 0.20.0 · 1m 10s · web · run artifacts}

🟢 e2e Electron stories (Haiku): 3 passed

3 agent steps · 3 model calls · 7.9k tokens (22% cached)
Spend: $0.02 of the $1.00 cap (proxy ledger)

All 3 tests in 3 files

	Test	Agent	Time
🟢	electron/deepLink.e2e.ts · 1 passed	1 step · 1 call	5.7s
🟢	a deep link of a second launch starts a chat in the running app	1 step · 1 call	5.7s
🟢	electron/menuBar.e2e.ts · 1 passed	1 step · 1 call	6.0s
🟢	the workspace menu bar is the desktop title bar	1 step · 1 call	6.0s
🟢	electron/smoke.e2e.ts · 1 passed	1 step · 1 call	5.0s
🟢	Xum Electron opens the playground workspace	1 step · 1 call	5.0s

_{e2e 0.20.0 · 18.7s · electron · run artifacts}

🟢 e2e agent repros (Haiku): no tests selected

Spend: $0.00 of the $1.00 cap (proxy ledger)

_{e2e 0.20.0 · 5.3s · 2 targets (web, phone) · run artifacts}

[ThomasK33]

VM A/B rule for this PR (perf owner, workspace 8e18f4aada), declared before any data.

Design: same-session VM A/B, 6+6 seeded runs per page and device (ABBA order), base = main f446f78738, head = this PR e65951282d. Statistic: median.

GOOD (merge) only if all hold:

1. Seeded mobile LCP median, head <= 2,058.2 ms (VM reference 1,958.2 + 100), and late runs (LCP > 2,258.2) <= 1 of 6.
2. Seeded mobile performance score median, head >= 90.
3. Seeded desktop LCP median, head - base <= +20 ms.
4. Seeded mobile TBT median, head - base <= +25 ms.
5. First-run mobile and desktop score medians, head >= base - 1.

BAD (no merge): desktop LCP head - base > +40 ms, or mobile TBT head - base > +50 ms with the same long task growing, or mobile LCP median > 2,258.2 ms. Anything in between: one more 3+3 block, then judge on all runs.

Separately from the A/B: the 31 pending Pixel snapshots need a person to confirm that only Geist Mono text (code, inline code, terminal) moved back to its pre-#7625 line box.

---

Generated with xum • Model: anthropic:claude-opus-5-5 • Thinking: high

[ThomasK33]

Readiness record for head e65951282d (perf owner, workspace 8e18f4aada):

• CI: Required passes on this head. Pixel / Review (not required) has 31 snapshots waiting for a person: tracked below.
• Codex: 1 round, clean on e65951282d (👍). 0 open review threads.
• VM A/B (rule declared before data: issuecomment-6425877557), base main f446f78738, head e65951282d, ABBA 6+6, all runs quiet. I read the medians from the report files:
  • Seeded mobile: LCP 8,570.5 -> 1,959.0 ms, late runs 0/6, score 69.5 -> 93.5, TBT 279 -> 275 ms. Clauses 1, 2, 4 pass.
  • Seeded desktop: LCP 1,655.0 -> 1,654.5 ms. Clause 3 passes.
  • First-run mobile score 98.0 -> 97.5, desktop 100 -> 100. Clause 5 passes. No BAD clause holds.
• Main moved to 97d86d75a5 (🤖 fix: drop remote plugin snapshots when project trust changes outside the app #7640, src/node only). The merge queue tests the merged tree.
• Independent advisor: font values checked against geist 1.5.1's GeistMono-Regular.ttf (955/-245, unitsPerEm 1000, lineGap 0, USE_TYPO_METRICS set). It asked for the Pixel review and a rebase. I declined the rebase (the merge queue tests the merged tree, and the delta is node-only). The Pixel review is a follow-up: the change puts back the Geist Mono line box main had until 🤖 fix: bump effect to stable 4.0.4 and dependencies within range #7625. If a snapshot shows any other change, I revert this PR.

Decision: ready with a tracked follow-up (Pixel snapshot review by a maintainer).

---

Generated with xum • Model: anthropic:claude-opus-5-5 • Thinking: high