fix: repeat copy-mode navigation keys

[ogulcancelik]

Summary

• repeat all copy-mode cursor and scroll navigation when terminals report explicit repeat events, including arrows, Page Up/Down, Home/End, Vim-style motions, and Ctrl-B/F/U/D
• keep confirm, exit, prefix, selection, search-opening, and other modal actions single-shot
• preserve input-source ownership with a copy-navigation-specific consumed lease

Reproduction

macOS Ghostty reports held keys as Kitty Repeat events; Windows can report repeated native presses. Before this change, those paths moved once in copy mode. The patched build repeats navigation continuously without allowing action-key repeats to escape the mode.

Tests

• just check
• automated macOS Ghostty → SSH validation: held Up changed copy-mode scroll offset from 0 to 53
• regression coverage for Ghostty enhanced arrows, Windows native character repeats, Ctrl navigation, both input routers, and action-key suppression

Refs #2371

[coderabbitai]

📝 Walkthrough

Walkthrough

Copy-mode navigation now supports explicit terminal key-repeat events for arrows, Page Up/Down, Home, and End. Dedicated input leases route these events through copy-mode handling while preserving existing repeat behavior for other keys.

Changes

Copy-navigation repeat handling

Layer / File(s)	Summary
Navigation repeat contract and test src/app/input/copy_mode.rs	AppState permits repeats only for selected navigation keys in copy mode. The test validates repeated arrow movement and cursor clamping.
Repeat lease disposition src/app/input/lease.rs	ConsumedInputLease records copy-navigation repeats. The lease table identifies these leases and excludes them from generic repeat planning.
Press and repeat event routing src/app/mod.rs, src/app/runtime.rs, docs/next/CHANGELOG.md	Input routing records copy-navigation leases on key presses and replays repeat events while copy mode remains active. The changelog documents the supported keys.

Estimated code review effort: 3 (Moderate) | ~20 minutes

Sequence Diagram(s)

sequenceDiagram
  participant Terminal
  participant AppState
  participant InputLeaseTable
  participant CopyModeHandler
  Terminal->>AppState: Send key press
  AppState->>InputLeaseTable: Record RepeatCopyNavigation
  Terminal->>AppState: Send KeyEventKind::Repeat
  AppState->>InputLeaseTable: Check copy-navigation lease
  AppState->>CopyModeHandler: Replay navigation repeat_count times

Loading

Possibly related issues

• Copy mode: allow held-key repeat for cursor-movement keys (keep suppressing confirm/exit repeat) #2371 — Covers the same copy-mode key-repeat behavior and repeat suppression paths.

Possibly related PRs

• herdrdev/herdr#2142 — Provides the input-lease repeat-routing system extended by this change.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly and concisely describes the main change: repeating copy-mode navigation keys.
Description check	✅ Passed	The description directly explains the copy-mode repeat behavior, preserved single-shot actions, implementation approach, reproduction, and validation.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/copy-mode-arrow-repeat

---

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

🧹 Nitpick comments (1)

src/app/input/copy_mode.rs (1)

1187-1199: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick win

Add coverage for both repeat routers.

This test exercises App::route_client_input, which covers the headless routing path. The PR also changes src/app/runtime.rs, and the allowlist includes PageUp, PageDown, Home, and End. Add one runtime-path test and cases for the remaining allowlisted keys.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: e08b7c75-531f-4917-9e1a-26f5e9b923ae

📥 Commits

Reviewing files that changed from the base of the PR and between eb93ec8 and 3638cce.

📒 Files selected for processing (5)

• docs/next/CHANGELOG.md
• src/app/input/copy_mode.rs
• src/app/input/lease.rs
• src/app/mod.rs
• src/app/runtime.rs

[ogulcancelik]

@coderabbitai review

[coderabbitai]

⚠️ Action not completed

Already reviewed.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.

[joonhwan]

Though I'm interested in this fix , when i cargo run this branch, copy mode's keys(i.e : j,k,ctrl-u,ctrl-d ...) cannot be repeated(For Windows herdr test).

[ogulcancelik]

@joonhwan Thanks for catching this you were right. The original allowlist covered only non-character navigation. 7cce478 now includes h/j/k/l, word and paragraph motions, Ctrl-B/F/U/D, and the other safe copy-mode movements, with a Windows native-repeat regression test. Could you confirm, exit, prefix, selection, and search-opening actions remain single-shot?

[greptile-apps]

Greptile Summary

The PR makes explicit repeat events replay copy-mode navigation while preserving single-shot modal actions and input-source ownership.

• Adds a copy-navigation repeat allowlist covering arrows, paging, line edges, search traversal, and Vim-style motions.
• Adds a dedicated consumed input lease and equivalent replay behavior for local and headless/client input paths.
• Adds regression tests for Kitty/Ghostty events, Windows character navigation, control-key paging, search prompts, and prefix handling.
• Documents the fix in the next-release changelog.

Confidence Score: 5/5

The PR appears safe to merge with no actionable correctness or security issues identified.

The local and headless paths consistently establish, replay, and release copy-navigation leases, while the allowlist and mode guards keep non-navigation actions single-shot.

Important Files Changed

Filename	Overview
src/app/input/copy_mode.rs	Adds a tightly scoped repeatability predicate and tests covering navigation, modal-action exclusions, search prompts, prefixes, and platform-specific repeat events.
src/app/input/lease.rs	Adds a dedicated consumed-lease disposition used to retain copy-navigation ownership across explicit repeat events.
src/app/mod.rs	Extends headless/client event routing to establish copy-navigation leases on presses and replay matching repeat events.
src/app/runtime.rs	Mirrors the copy-navigation lease and replay behavior in the asynchronous local-input runtime path.
docs/next/CHANGELOG.md	Documents repeat support for copy-mode navigation and continued single-shot handling for other actions.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Raw key event] --> B[Normalize against input lease]
  B --> C{Press or Repeat?}
  C -->|Press| D{Copy-mode navigation?}
  D -->|Yes| E[Execute navigation once]
  E --> F[Store RepeatCopyNavigation lease]
  D -->|No| G[Use normal input routing]
  C -->|Repeat| H{Matching copy-navigation lease and Copy mode active?}
  H -->|Yes| I[Replay navigation repeat_count times]
  H -->|No| J[Use normal repeat plan]
  K[Key release] --> L[Remove lease]

Loading

_{Reviews (1): Last reviewed commit: "fix: repeat copy-mode navigation keys" | Re-trigger Greptile}

[joonhwan]

@ogulcancelik

@joonhwan Thanks for catching this you were right. The original allowlist covered only non-character navigation. 7cce478 now includes h/j/k/l, word and paragraph motions, Ctrl-B/F/U/D, and the other safe copy-mode movements, with a Windows native-repeat regression test. Could you confirm, exit, prefix, selection, and search-opening actions remain single-shot?

This feature now works in windows 11 / wt.exe ! The test results follow.

• repeat check
  • "char movemetn" - h/j/k/l : ok
  • "page up" - Ctrl+Shift+b(Ctrl-B) : ok (Ctrl+b is prefix in my case)
  • "page down" - Ctrl+Shift+f(Ctrl-F) or Ctrl+f(Ctrl-f) repeat : ok 🫨
  • "half-page up/down" - Ctrl+d/u or Ctrl+Shift+d/u repeat : ok
• no repeat check
  • "search" - "/" or "?" : ok
  • "visual selection" - "v/space" : ok
  • "yank" - "y/enter" : ok. [1]
  • "exit cursor mode" - "q/esc", : ok. [1]

[1] : no input from repeating keys after exiting cursor mode, which is qood

BTW, since I've been using herdr, i've not have any chance to use "n/N" keys successfully(like. 10 h) . 😢

[shibayu36]

Thank you for working on the copy-mode key repeat! This behavior has bothered me too, so I'm glad to see a fix in progress.

One suggestion: could the repeat decision be based on what the key press actually did, instead of a list of keys?

As analyzed in #2371, what this suppression needs to prevent is: a key triggers an action that leaves the mode, the key is still held, and the remaining repeats leak into the next mode as input. If so, whether a held key is safe to repeat is decided by the action it triggered, not by which key it is bound to.

Currently is_repeatable_navigation_key is a list of keys that are allowed to repeat, and its char list (h/j/k/l, g, G, w, b, e, …) duplicates the dispatch list in handle_copy_mode_key. Adding a new navigation motion later means updating both places, and missing one silently leaves only the new key non-repeating, which is hard to notice. If instead the repeat was stopped only when the press is observed to do something that should stop it — such as leaving the mode — everything else would repeat automatically, and the allowlist would not need updating when navigation motions are added.

One more reason: if copy-mode keybindings become configurable in the future (I proposed this in #587), a hardcoded key list stops working — for example, binding cursor-up to ctrl+p would leave it non-repeating. Deciding from the result of the press keeps working no matter how keys are rebound.