fix(web,desktop,mobile): QA sweep — naming, titlebar, nav, tooltips, terminal

[maarco]

QA pass over the running desktop app. One commit rather than several: the pre-commit hook re-stages the whole working tree, so splitting by concern was not achievable without disabling it.

Branding

(Alpha) dropped from the product name across desktop, web and mobile. Renaming the bundle moves Electron's userData directory, so legacyUserDataDirName becomes a list (T3 Code (Alpha), Marcode (Alpha)) probed concurrently — probing sequentially added enough filesystem latency for Electron's ready event to beat protocol.registerSchemesAsPrivileged, crashing the packaged app on launch. Mocked filesystems make sequential and concurrent probes indistinguishable, which is why the suite could not see it.

Titlebar

Settings and usage rendered their breadcrumb under the macOS traffic lights. Two causes: the inset class keys off a data-sidebar-state ancestor and Marcode mounts no sidebar there; and --workspace-titlebar-content-left was declared only on the sidebar wrapper. Variable moves to :root, sidebarless surfaces get an unconditional inset.

Navigation

• Removes the dead /settings/beta entry (upstream deleted that page with sidebar v2).
• Moves Pull Requests from Settings into the home category — getActiveCategory already resolved the route there, so the wrong pill highlighted. Usage stays under Settings.
• Pill could run past the window edge: no-scrollbar was referenced twice and defined nowhere, and the cap was a flat max-w-[100vw]. Cap now derives from the pill's own dock offset and scale, on both axes.

Tooltips

14 icon-only controls across the pill nav, pull-request surfaces and usage page had no hover description.

Open-in picker

Popover used Base UI's default closeDelay of 0, so the pointer could not cross the trigger gap. Rows gained the watermark treatment from PillNavCard.

Terminal

xterm.js was imported without its stylesheet, anywhere. The rule hiding .xterm-char-measure-element never applied, so xterm's char-width scratchpad rendered its 32-repeat probe strings as visible text above the first prompt — the reported 333...MMM garbage. Never terminal data, which is why the PTY, sanitiser, history files and websocket frames were all clean. Separately, trimBufferToBytes was UTF-8-safe but ANSI-blind and could cut mid-sequence; now cuts only at escape-sequence boundaries.

Typography

10 text classes reconciled to Marcode's scale. Four trace to upstream commits that merged cleanly and were never reconciled.

Also adds .marcode to .gitignore — only upstream's .t3 was listed.

🤖 Generated with Claude Code

https://claude.ai/code/session_01EwikZhkf3NAa9KQkULRSre

[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	11.3 KiB	11.3 KiB	−18 B (−0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	5.5 KiB	5.5 KiB	+7 B (+0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	5.9 KiB	5.9 KiB	−25 B (−0.4%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	49.7 KiB	49.7 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	16	16	0 (0.0%)	21	✅
Claude	Total thread wire	11.3 KiB	11.3 KiB	+28 B (+0.2%)	15.1 KiB	✅
Claude	Thread snapshot wire	5.5 KiB	5.5 KiB	+1 B (+0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	5.8 KiB	5.9 KiB	+27 B (+0.5%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	50.6 KiB	50.6 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	16	16	0 (0.0%)	21	✅

Baseline: 7c0314e · PR result: e25a4f2 · 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: 94.6 KiB
• Claude decoded thread snapshot: 95.4 KiB

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