v0.9.0: Unicode, CJK, and terminal-width rendering QA

[Hmbown]

Problem

Several reports point at text rendering problems that are broader than a single widget: CJK text, mixed-width Unicode, wrapped output, selector rows, and terminal-width changes can break readability or make output look incomplete.

This should be handled as an explicit QA lane after the v0.8.66 visual matrix starts landing, rather than as isolated fixes per report.

Related reports and themes:

• garbled Chinese / CJK real-time output reports
• incomplete displayed output reports
• selector and settings surfaces with width/contrast issues
• terminal-native copy including visual line breaks from wrapped output (TUI: terminal-native copy includes visual line breaks from wrapped output #1853)
• narrow Windows/macOS terminal rendering differences

Scope

Add Unicode and terminal-width QA coverage for CodeWhale TUI surfaces.

Cover:

• CJK task titles and transcript output
• mixed English/CJK rows in sidebars, modals, and selectors
• wide glyphs, combining marks, emoji, and ambiguous-width punctuation where practical
• copy/export behavior for wrapped output
• terminal resize/narrow-width behavior
• statusline and modal title truncation

Suggested Implementation

• Add fixture strings with CJK, mixed-width Unicode, and long branch/objective names.
• Test width calculations through the same helpers used by production rendering.
• Add snapshot/golden-state cases for narrow and medium terminals.
• Ensure truncation preserves semantic prefixes/suffixes and never splits grapheme clusters.
• Audit copy behavior separately from visual wrapping.

Acceptance Criteria

• CJK and mixed-width rows do not corrupt borders or hide key state.
• Truncation/wrapping respects display width, not byte length.
• Copy/export can preserve logical text without visual wrap artifacts where promised.
• At least one macOS Terminal-like and one Windows/PowerShell-like rendering scenario is represented by fixtures or documented manual QA.
• Known related issues are linked or moved into this lane as they are reproduced.

[Hmbown]

Additional screenshot-derived Unicode/CJK case from 2026-06-23 dogfood:

The sub-agent overlay includes a running agent named 抹香鲸 next to English metadata:

• 抹香鲸 agent_e0b2dcf1 running 10+ 124838ms
• role: implementer
• git: branch codex/issue-3439-zhipu-glm-fixture @ issue-3439
• objective: QUESTION: Add Zhipu GLM as a first-class provider-scoped ...

This is a useful fixture because the CJK display name sits in the same table-like row as ASCII IDs, numeric columns, and status text. The current visual result looks fragile: columns do not read as stable, and it is hard to tell whether the row layout is using display width, byte length, or ad hoc string truncation.

Please include a focused case for:

• CJK agent/task display names in sub-agent rows.
• Mixed English/CJK rows with status, ID, step count, elapsed time, branch, and objective lines.
• Narrow/medium terminal widths where CJK names should not shift status columns, corrupt borders, or hide the running/completed state.
• Grapheme/display-width-aware truncation for objectives that include provider names, issue numbers, branch names, punctuation, and mixed-language text.

Acceptance addition: a CJK display name should not make adjacent status/time columns visually drift, and truncation should happen at semantic boundaries where possible rather than splitting the row into hard-to-read fragments.