fix(tui): truncate the spinner label by display width, not rune count

[dennisonbertram]

Closes #1432

The bug

shortenLabel exists to guarantee one thing: at narrow widths the label yields so the (esc to interrupt) hint survives, because the hint is the only actionable part of that line. For double-width characters it did the opposite.

width=40  "· Waiting for 模型模型模型… (esc to inte"
width=40  "· Running 🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀… (esc to "
width=50  "· Waiting for 模型模型模型模型模型模型模型模型 (es"

Cause

Two units, mixed. The budget was computed in display columns:

budget := width - lipgloss.Width(glyph) - lipgloss.Width(CancelHint) - 2

and the truncation was applied in runes:

runes := []rune(label)
if len(runes) > budget {
    label = string(runes[:budget-1]) + "…"
}

CJK and emoji occupy two columns per rune, so a rune-budgeted label could be twice its allowance in columns. The line overran, and the final MaxWidth clamp trimmed from the right — taking the hint.

The rendered width was never wrong, which is exactly why this hid. The clamp always produced a correctly-sized line; it just produced the wrong one.

Fix

truncateToWidth accumulates lipgloss.Width per rune until the budget is reached, counting the ellipsis's own width.

Deliberately not grapheme-cluster aware: a combining mark or ZWJ emoji sequence can still be split. That is a strict improvement over rune counting and a far smaller change than full segmentation, which stays out of scope until a real case appears rather than being built speculatively.

After

w=40  "· Waiting for 模型模… (esc to interrupt)"
w=50  "· Waiting for 模型模型模型模型… (esc to interrupt)"
w=40  "· Running 🚀🚀🚀🚀🚀… (esc to interrupt)"
w=40  "· Waiting for some-l… (esc to interrupt)"
w=40  "· Running bash (esc to interrupt)"          (unchanged, fits)

Verification

TestSpinnerKeepsCancelHintWithWideRunes — CJK, emoji, and an ASCII control at widths 40 and 50. Red first on both wide cases:

width 40: cancel hint truncated, got "· Waiting for 模型模型模型… (esc to inte"
width 40: cancel hint truncated, got "· Running 🚀🚀🚀🚀🚀🚀🚀🚀🚀🚀… (esc to "

It asserts the hint's survival, not the line's width. Asserting width is what let this hide — the width was always right. The ASCII control guards the other direction: a fix that satisfied the hint assertion by over-truncating every label would show up there as a needlessly stunted rendering.

Full TUI suite green under -race. Rendered output inspected directly, not only asserted.

How this was found

By gpt-6-astra through the Surplus proxy — an untrusted external reviewer handed the packed source with no tools and no filesystem access — then confirmed locally by rendering the cases before believing any of it.

Our own tests were green throughout and stayed green, because they were ASCII-only. The code asserted a guarantee in a comment that it did not keep, and an outside reader that could not run anything caught what the suite structurally could not.

Honest signal-to-noise from that review: three findings, one confirmed (this), one unconfirmed and overlapping with it (a claim that MaxWidth may wrap rather than clip — not observed), one false positive (elapsed-time nondeterminism in snapshots, which set startTime explicitly).

🤖 Generated with Claude Code

https://claude.ai/code/session_01WJGxhoFhA8JjkwZFcLGdS5

[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 2026-09-08T17:46:58.622331Z	1fe3b02	PR opened

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

Here are some automated review suggestions for this pull request.

Reviewed commit: 1fe3b02af1

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]

Make the ASCII control assert retained label text

This control does not detect the over-truncation it claims to guard against: an implementation that reduces every overlong label to only … (or drops it entirely) still passes because the shared assertions check only that the hint survives and the result fits. Assert an expected retained ASCII prefix or minimum label width so this case actually distinguishes correct truncation from needless label loss.

AGENTS.md reference: AGENTS.md:L31-L31

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Truncate using whole grapheme display widths

For multi-rune emoji such as ❤️ or 1️⃣, display width is not additive per rune: the variation-selector/keycap runes measure as zero alone, while the completed grapheme occupies two columns. This loop therefore undercharges these labels, can return a value wider than budget, and causes View's final MaxWidth clamp to truncate the cancel hint again at narrow widths. Iterate grapheme clusters (or use an ANSI/grapheme-aware width truncator) so the accumulated width matches lipgloss.Width of the returned prefix.

Useful? React with 👍 / 👎.