TUI: skip unchanged running-command height remeasurement

[harryalbert]

Description

Adds a targeted follow-up for expanded shell commands that are still running inside Warp Agent CLI blocks. The downstack paint fix (#14594) bounds inline terminal rendering to visible rows, but the transcript viewport still remeasured the entire agent block on every layout while an expanded command was running—even when the terminal row extent had not changed. Generic streaming status also forced repeated measurement despite agent-output updates already dirtying their rich-content item directly.

This change records each running shell command's rendered height and displayed command/output row extents after layout. Stable-width viewport layouts now reuse the cached agent-block height while those extents are unchanged; new terminal rows, width changes, disclosure changes, and normal dirty invalidations still trigger measurement. The viewport collects candidate rich-content IDs while holding the terminal-model lock, then releases it before querying child views, avoiding nested terminal-model locking when reading live command geometry. Regression coverage verifies same-row terminal updates reuse the cached height, row growth remeasures, completion stops dynamic measurement, and ordinary agent-output updates continue refreshing canonical scroll extent.

Benchmark at 120×50 versus downstack PR #14594. The fixture invalidates only the transcript viewport—not the retained agent block—so it measures the unchanged-row relayout path affected by this PR:

• Expanded running command, 100 output rows: 19.786 → 18.417 µs/frame (7.1% faster, p=0.01).
• Expanded running command, 1,000 output rows: 19.119 → 18.472 µs/frame (3.9% faster, p=0.02).

Agent conversation: https://staging.warp.dev/conversation/963ead54-9843-4cc1-8d8b-dbf695ee7046

Linked Issue

N/A — Warp Agent CLI transcript rendering performance.

Testing

• cargo nextest run -p warpui_core --features tui — 557 passed, 7 skipped.

• cargo nextest run -p warp_tui — 937 passed.

• cargo clippy -p warp_tui --features test-util --benches -- -D warnings

• cargo bench -p warp_tui --features test-util --bench transcript_bench -- 'tui_transcript/running_agent_command/viewport_relayout_unchanged_frame'

• ./script/format

• I have manually tested my changes locally with ./script/run-tui

Agent Mode

• Warp Agent Mode - This PR was created via Warp's AI Agent Mode

CHANGELOG-NONE

Co-Authored-By: Oz oz-agent@warp.dev

[harryalbert]

• TUI: skip unchanged running-command height remeasurement #14595 👈 (View in Graphite)
• TUI: clip inline terminal block painting #14594
• [3/4] TUI: reuse retained text measurements #14585 : 1 other dependent PR (#14586 )
• [2/4] TUI: bound clipped viewport painting #14582
• [1/4] TUI: add retained transcript performance benchmarks #14575
• master

This stack of pull requests is managed by Graphite. Learn more about stacking.

[oz-for-oss]

@harryalbert

I'm starting a first review of this pull request.

You can view the conversation on Warp.

I completed the review and no human review was requested for this pull request.

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

Powered by Oz

[oz-for-oss]

Overview

This PR narrows TUI agent-block height polling so expanded running command blocks are remeasured only when their terminal row extent changes, with supporting tests and a benchmark fixture.

Concerns

• This is a user-visible TUI behavior change, but the PR description does not include manual testing output or acceptable TUI evidence such as a terminal transcript, render-to-lines/TuiBuffer snapshot diff, or ./script/run-tui capture demonstrating the running-command row-growth behavior end to end.

Verdict

Found: 0 critical, 1 important, 0 suggestions

Request changes

Comment /oz-review on this pull request to retrigger a review (up to 3 times on the same pull request).

Powered by Oz