perf(web): dedupe terminal mouse motion reports

[t3-code]

summary

• stop forwarding duplicate terminal mouse-motion packets when the pointer stays inside the same terminal cell
• keep all meaningful motion, button, modifier, press, and release changes intact
• reset motion deduplication after pointer leave and non-motion mouse actions
• add regression coverage for the dedupe/reset behavior

root cause

OpenCode v2 enables DEC 1003 any-event mouse tracking. Browser pointer events use pixel coordinates, while the terminal protocol reports cell coordinates. A normal pointer sweep therefore produced several byte-identical SGR motion packets per cell, and T3 Code sent every duplicate through the websocket and PTY.

The fix drops only consecutive byte-identical motion packets. Absolute cell changes still send, while press/release events always send and reset the motion baseline.

benchmark

OpenCode v2 0.0.0-beta-17823 running in T3 Code's real in-app terminal, with 1,000 pointer events over the same sweep path. Values are medians from five runs.

• websocket frames: 1,006 → 188, 81.3% fewer
• websocket payload: 263,276 → 48,425 bytes, 81.6% less
• browser task time: 1,381.9 → 861.9 ms, 37.6% less
• script time: 724.3 → 327.3 ms, 54.8% less

tested in the real app

OpenCode v2 running in the isolated T3 Code dev client's terminal after the fix. Its full-screen TUI, mouse tracking, input, and rendering remained intact.

verification

• 69 Ghostty terminal unit tests pass
• web typecheck passes
• web production build passes
• focused lint and format checks pass
• independent OpenCode v2 review found no logic or security blockers
• benchmark rerun against the live in-app terminal

built with gpt-5.6-sol via hermes agent.

request provenance

• requested by: @maria-rcks

---

Note

Medium Risk
Changes what mouse sequences reach the PTY for apps using any-event tracking. Duplicate same-cell motion is dropped by design; incorrect baseline resets could skip a legitimate first motion after a mode change.

Overview
Stops forwarding duplicate SGR mouse-motion packets when the pointer stays in the same terminal cell under DEC 1003 any-event tracking. Press/release still always send, and they (plus pointer leave, non-reporting motion, session reset, and DEC 1003 mode flips) reset the motion baseline so the next cell report is not dropped.

sendMouse now gates onData through resolveTerminalMouseData. Writes and pointer moves also sync cached any-event tracking so a mode change without pointer movement starts a fresh session. Unit tests cover the dedupe and tracking-reset helpers.

^{Reviewed by Cursor Bugbot for commit 9782d0f. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Dedupe terminal mouse motion reports in GhosttyTerminalSurface

• Adds resolveTerminalMouseData to suppress outgoing motion reports when the encoded mouse data matches the previous motion baseline; non-motion actions (press/release) reset the baseline
• Adds resolveTerminalMouseTrackingState to clear the stored motion baseline when the any-event tracking mode toggles
• onPointerMove, onPointerLeave, write, and resetAndWrite now synchronize the cached tracking mode and clear the motion baseline at the appropriate points
• Behavioral Change: sendMouse only emits data when resolveTerminalMouseData.send is true, so duplicate same-cell motion events are no longer forwarded to the terminal; lastMouseMotionData is the new private field controlling this dedup state

^{Macroscope summarized 9782d0f.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.4 KiB	13.4 KiB	+1 B (+0.0%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	+2 B (+0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.5 KiB	−1 B (−0.0%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.0 KiB	55.0 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	13.4 KiB	13.4 KiB	−12 B (−0.1%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	−10 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.6 KiB	−2 B (−0.0%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	55.8 KiB	55.8 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: c3e3709 · PR result: 9782d0f · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 5bcb285. Configure here.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This performance optimization changes terminal mouse event handling by filtering duplicate motion reports. While well-tested, it modifies what data is sent to the terminal, which is a behavioral change to core input handling that warrants human review.

^{You can add or adjust custom eligibility rules. Learn more.}