feat(input): a Toggle for Sprint on the Controls page, Off (held) by default #1977

Merged
HumanBean merged 10 commits from llun/Longshot:feat/sprint-toggle into main 2026-11-21 06:20:47 +00:00
Member

Summary

A follow-up to #1948 (Sprint held on L3, on the Controls tab), from a hand test of it: the Controls page's Toggle column showed - for Sprint. Sprint now has a Toggle, as the Switch's buttons do:

  • Off (the default): Sprint runs while its key or button is held, as #1948 made it.
  • On: a press of Sprint's key or button starts it and the next press ends it (sprint_toggle_step, host/nn_hid.c). Opening Insert, pressing Plus or Minus, or the window losing the focus ends it, as the buttons' toggles are ended.

It is saved as toggle_sprint= (opts.sprint_toggle: the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit) and read by the host as HOS_OPT_SPRINT_TOGGLE (98). Sprint's bindings stay rebindable in the Input 1, Input 2 and Keyboard columns; the Controls page's Reset turns the Toggle off with the others. The Xbox binding cells use L3 and R3 labels, Sprint appears after Walk, and binding help explains Sprint's inputs and when L3 moves to Sprint. The binding cells use L3 and R3 labels, including for Xbox prompts; Sprint follows Walk in the list, and the page gives more specific binding help.

Also: tests/test_walk_key.py failed on main after #1948 (its stand-ins left sprint_frame set but never read); it builds again. The Graphics guide repair supports the Reset HUD placement action and documents that Custom HUD controls appear when HUD layout is Custom and HOS_WIDE_UI does not override it. Website tests cover the visibility condition and reset action.

Read first: sprint_toggle_step and its call in npad_states (host/nn_hid.c), has_toggle/toggle_on/toggle_set in gpu/vk/vk_menu.cpp.

Testing

Tested: Windows 11, llvm-mingw: tests/test_sprint.py (sprint_toggle_step on its own: Toggle off passes held through; on, a press latches and the next unlatches; a reset ends it and a press held across the reset doesn't restart it; and the wiring: the Toggle column, toggle_sprint saved, the host's option), test_input_background, test_walk_key, test_upscaler_round_trip, test_menu_env_overrides, test_bound_keys, test_option_ids, test_menu_modules, test_menu_i18n, test_settings_table: all pass. test_pads_after_menu fails on main too (#1721's stand-in pad_count_now, unused). The game is built from this commit; it has not run in the game yet: opened at the user's request before the run.

The full suite and lint have not been run on the updated head; their results are pending. The earlier targeted run named test_pads_after_menu as a failure on main too.

Still to test:

  • Windows, a controller and the keyboard, by hand: the Controls page, Sprint's Toggle On: one press of L3 or Shift starts Sprint, the next ends it; Off: only while held. Rebind Sprint's pad button and key and check both still work (the user, by hand).
  • Linux: the build (@Beanstalk Linux build and start).

Measurements

Nothing to measure: with the Toggle off, one more compare per input read; #1929 measured Sprint against main within the spread.

Risks

  • Sprint Toggle is off by default. The Controls page still shows the new L3 and R3 labels, Sprint's position after Walk, and updated binding help. longshot_options.ini gains a toggle_sprint= line.
  • Reverting removes the Toggle; Sprint stays held-only.

Checklist

Ready for review

  • One topic, on current main; the title is a Conventional Commit.
  • Tested on my own PC, in the game for a change the game runs, as Testing says. (not run in the game yet: opened at the user's request; the user is testing it by hand)
  • mise run test passes (any failure that also fails on main is named). Full suite and lint still need to run on the updated head; test_pads_after_menu was reported failing on main (#1721's unused pad_count_now).
  • Measured against main, or Measurements says why there is nothing to measure.
  • No game files, build output or hand-edited generated files; third_party/ untouched; nothing GPL-incompatible.

When it applies

  • A check under tests/ that reads changed source is updated, not weakened.
  • Host code builds on Windows (llvm-mingw) and Linux, with UTF-8 paths. (built on Windows; Linux not built here)
  • Screenshots of the same spot on main and here, for a change that shows on screen.
  • Docs updated.
  • AI assistance recorded with Assisted-by: / Co-Authored-By: trailers.

Ready to merge

  • Everything in Still to test has run, its results in Tested.
  • Tier 1 hardware it can affect is tested (docs/dev/standards.md, "Hardware tiers").
  • Approved in review, every comment answered.
  • CI passes on the head; on current main, measured again or a comment says why the numbers still hold.

🤖 Generated with Claude Code

## Summary A follow-up to [#1948](https://git.magicbean.dev/MagicBeanDev/Longshot/pulls/1948) (Sprint held on L3, on the Controls tab), from a hand test of it: the Controls page's **Toggle** column showed `-` for Sprint. Sprint now has a Toggle, as the Switch's buttons do: - **Off** (the default): Sprint runs while its key or button is held, as #1948 made it. - **On**: a press of Sprint's key or button starts it and the next press ends it (`sprint_toggle_step`, `host/nn_hid.c`). Opening Insert, pressing Plus or Minus, or the window losing the focus ends it, as the buttons' toggles are ended. It is saved as `toggle_sprint=` (`opts.sprint_toggle`: the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit) and read by the host as `HOS_OPT_SPRINT_TOGGLE` (98). Sprint's bindings stay rebindable in the Input 1, Input 2 and Keyboard columns; the Controls page's Reset turns the Toggle off with the others. The Xbox binding cells use L3 and R3 labels, Sprint appears after Walk, and binding help explains Sprint's inputs and when L3 moves to Sprint. The binding cells use L3 and R3 labels, including for Xbox prompts; Sprint follows Walk in the list, and the page gives more specific binding help. Also: `tests/test_walk_key.py` failed on `main` after #1948 (its stand-ins left `sprint_frame` set but never read); it builds again. The Graphics guide repair supports the Reset HUD placement action and documents that Custom HUD controls appear when HUD layout is Custom and `HOS_WIDE_UI` does not override it. Website tests cover the visibility condition and reset action. Read first: `sprint_toggle_step` and its call in `npad_states` (`host/nn_hid.c`), `has_toggle`/`toggle_on`/`toggle_set` in `gpu/vk/vk_menu.cpp`. ## Testing **Tested:** Windows 11, llvm-mingw: `tests/test_sprint.py` (`sprint_toggle_step` on its own: Toggle off passes held through; on, a press latches and the next unlatches; a reset ends it and a press held across the reset doesn't restart it; and the wiring: the Toggle column, `toggle_sprint` saved, the host's option), `test_input_background`, `test_walk_key`, `test_upscaler_round_trip`, `test_menu_env_overrides`, `test_bound_keys`, `test_option_ids`, `test_menu_modules`, `test_menu_i18n`, `test_settings_table`: all pass. `test_pads_after_menu` fails on `main` too (#1721's stand-in `pad_count_now`, unused). The game is built from this commit; it has not run in the game yet: opened at the user's request before the run. The full suite and lint have not been run on the updated head; their results are pending. The earlier targeted run named `test_pads_after_menu` as a failure on `main` too. **Still to test:** - Windows, a controller and the keyboard, by hand: the Controls page, Sprint's Toggle On: one press of L3 or Shift starts Sprint, the next ends it; Off: only while held. Rebind Sprint's pad button and key and check both still work (the user, by hand). - Linux: the build (`@Beanstalk` Linux build and start). ## Measurements Nothing to measure: with the Toggle off, one more compare per input read; #1929 measured Sprint against `main` within the spread. ## Risks - Sprint Toggle is off by default. The Controls page still shows the new L3 and R3 labels, Sprint's position after Walk, and updated binding help. `longshot_options.ini` gains a `toggle_sprint=` line. - Reverting removes the Toggle; Sprint stays held-only. ## Checklist ### Ready for review - [x] One topic, on current `main`; the title is a Conventional Commit. - [ ] Tested on my own PC, in the game for a change the game runs, as Testing says. (not run in the game yet: opened at the user's request; the user is testing it by hand) - [ ] `mise run test` passes (any failure that also fails on `main` is named). Full suite and lint still need to run on the updated head; `test_pads_after_menu` was reported failing on `main` (#1721's unused `pad_count_now`). - [x] Measured against `main`, or Measurements says why there is nothing to measure. - [x] No game files, build output or hand-edited generated files; `third_party/` untouched; nothing GPL-incompatible. ### When it applies - [x] A check under `tests/` that reads changed source is updated, not weakened. - [ ] Host code builds on Windows (llvm-mingw) and Linux, with UTF-8 paths. (built on Windows; Linux not built here) - [ ] Screenshots of the same spot on `main` and here, for a change that shows on screen. - [x] Docs updated. - [x] AI assistance recorded with `Assisted-by:` / `Co-Authored-By:` trailers. ### Ready to merge - [ ] Everything in Still to test has run, its results in Tested. - [ ] Tier 1 hardware it can affect is tested (docs/dev/standards.md, "Hardware tiers"). - [ ] Approved in review, every comment answered. - [ ] CI passes on the head; on current `main`, measured again or a comment says why the numbers still hold. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
feat(input): Sprint has a Toggle on the Controls page, Off (held) by default
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 8s
PR description / Description follows the template (pull_request_target) Successful in 11s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 11s
CI / Workflow lint (actionlint) (pull_request) Successful in 11s
Third-party parts / As their notes say (pull_request) Successful in 12s
Website / Website on Cloudflare (pull_request_target) Has been skipped
CI / Website build and checks (pull_request) Has been cancelled
f73c817a5b
From the user's hand test of #1948: the Controls page's Toggle column showed "-" for Sprint. It now has one, as the
Switch's buttons do: Off, the default, Sprint while its key or button is held; On, a press starts it and the next ends it
(host/nn_hid.c sprint_toggle_step), ended as the buttons' toggles are (opening Insert, Plus or Minus, the focus lost).
Saved as toggle_sprint= (opts.sprint_toggle: Sprint's index is past opts.hold_toggle's bits), read by the host as
HOS_OPT_SPRINT_TOGGLE (98). Its bindings stay rebindable in the Input 1, Input 2 and Keyboard columns.

Also from #1948's merge: tests/test_walk_key.py's stand-ins read sprint_frame, which read_input sets and npad_states
(not cut there) reads, so the check builds again. tests: test_sprint (sprint_toggle_step on its own, the wiring),
test_input_background, test_upscaler_round_trip and test_menu_env_overrides follow. The user guide says it.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Member

Possibly related, for context:

  • issue #1881 Idea: Add an auto-run toggle like Hookshot — Open request for an auto-run toggle; this adds persistent Sprint behavior.
  • PR #613 WIP: feat(input): separate keyboard actions and binding pages — Open input PR also adds keyboard Sprint toggle behavior and touches Controls/input code.
Possibly related, for context: - issue #1881 Idea: Add an auto-run toggle like Hookshot — Open request for an auto-run toggle; this adds persistent Sprint behavior. - PR #613 WIP: feat(input): separate keyboard actions and binding pages — Open input PR also adds keyboard Sprint toggle behavior and touches Controls/input code.
BeanBot requested changes 2026-11-21 02:25:18 +00:00
Dismissed
BeanBot left a comment

Recommendation: Request changes. Sprint can latch again when focus returns, carry a note-button press into gameplay, or inject Sprint into an exclusive test run. The renderer also shares the new setting with the game thread without synchronization, and three existing test fixtures need updates.

Adds a saved Toggle cell to Sprint’s Controls-page binding. Off preserves hold-to-sprint, and On lets players take their finger off the key or stick click while running. The UI follows the existing binding grid, but the new input latch needs to respect menu, focus, note, and automation boundaries.

@fix will do (16)

  1. L1 gpu/vk/vk_replay.cpp:7739: The game thread reads opts.sprint_toggle through rx_option while the present thread can write it from Controls. (more on the line)
  2. L2 host/nn_hid.c:1472: During Insert or focus loss, read_input hides the held Sprint key, so the reset records was = 0. (more on the line)
  3. L3 host/nn_hid.c:1474: The note-wait check happens after the latch has already advanced. (more on the line)
  4. L4 host/nn_hid.c:1431: Script and bridge SPRINT now feed the player’s latch. A timed SPRINT+LUP followed by plain LUP keeps Sprint on after the timed press ends, contrary to the documented held input. An exclusive bridge movement can also inherit a latch set by a physical press before the request. Keep synthetic Sprint as a timed hold and suppress the player latch’s output while physical input is excluded.
  5. L5 (minor) host/nn_hid.c:1468: The latch only observes Sprint being disabled when a poll sees sprint_enabled false. (more on the line)
  6. L6 tests/test_menu_keys.py:343: The compiled fixture extracts the new load block without has_toggle or toggle_set, so compilation fails. Its checks at lines 356 and 358 also require the old inline predicates. Include the real helpers and keep coverage that Plus and Minus cannot be loaded, drawn, or selected as toggleable.
  7. L7 tests/test_crt_options.py:370: The compiled CRT fixture extracts save and the options loader without their new toggle_on, has_toggle, and toggle_set dependencies. The fixture fails to compile even though it only exercises CRT settings. Include the toggle helpers in its extracted code.
  8. L8 tests/test_menu_save_atomic.py:386: The Linux atomic-save fixture extracts save() without toggle_on. Its generated C++ fails to compile before exercising any disk-failure case. Include the getter and its Sprint setting in the fixture.
  9. L9 tests/test_sprint.py:491: The new wiring test matches source text, so removing the reset assignment or breaking the loader still passes. (more on the line)
  10. L10 (minor) tests/test_walk_key.py:336: The updated Walk fixture still ignores CC and can skip on a machine whose configured compiler has a different executable name or lives outside PATH. Reuse the Sprint fixture’s compiler selection. Apply the same selection to the default for the updated input-background fixture, whose suite declaration currently forces gcc.
  11. L11 (minor) gpu/vk/vk_menu.cpp:3709: The new saved key is accepted through atoi. (more on the line)
  12. L12 (minor) docs/user/graphics-and-controls.md:266: The main Sprint instructions and settings-table entry still say to hold the input and release it to stop. The Controls-tab Sprint help and the binding-cell help make the same unconditional claim. A player reading those after enabling Toggle gets the wrong instructions. Qualify the hold behavior as the default, point to the Toggle cell, and explain that the parent Sprint setting must be On.
  13. L13 (minor) gpu/vk/vk_menu.cpp:2559: The new whole-string help has no matching translation, so a player using a translated Controls page gets the Sprint Toggle instructions in English. (more on the line)
  14. L14 (minor) gpu/vk/vk_menu.cpp:1396: HOS_KEY_SPRINT is 31, which is the sign-bit position of a 32-bit int. (more on the line)
  15. L15 (minor): The PR ticks mise run test but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell.
  16. L17 (its part an edit can do) host/nn_hid.c:1468: Publish the toggle mode with the existing Sprint input snapshot and avoid a separate renderer option dispatch on each poll.

For the author (2) (a run, a measurement or a decision)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1468: The Measurements section calls this one extra compare, but with Sprint enabled and Toggle Off each poll also calls the renderer option dispatcher and updates edge state. (more on the line)

Checks: Still running: PR description / Description follows the template (pull_request_target), CI / Conventional Commit title (pull_request), CI / Workflow lint (actionlint) (pull_request), CI / Website build and checks (pull_request), Guide screenshots / Guide images and local UI source review (pull_request), Third-party parts / As their notes say (pull_request).

More: does it help, other systems, switches, what was read

Does it help: Unproven in game. The author reports a Windows 11 llvm-mingw build and passing targeted fixtures. The isolated latch test covers press, release, and reset with an input still visible to the helper. There is no physical-input run on this head, Linux build or start, or frame-time comparison against the base. The #1929 measurements predate this toggle.

Other systems: Windows and Linux share the changed Npad polling path, including Steam Deck controller input. The pad and keyboard binding formats and existing Sprint A-press handoff are preserved in the diff. The new latch affects ocarina input and scripted or bridge-driven runs as detailed below. The renderer change is confined to the menu, settings and option callback. No render pass, save-game format, audio path, GPU capability check, or build configuration changes.

Switches: No environment switches added or changed. The player setting toggle_sprint defaults to 0 (Off), which preserves held Sprint for existing settings files. HOS_OPT_SPRINT_TOGGLE is the host option ID for that setting.

Reviewed: f73c817a5b against 6e1521a70a: Read AGENTS.md, CONTRIBUTING.md, the Forgejo template, developer standards, the sole commit, the full eleven-file diff twice, and the surrounding code at each change. Followed the input and focus callers, binding snapshot, menu save/load and rollback, renderer option callback, and dependent test fixtures. Skimmed the guide manifest and translation tables. Unrelated portions of the large renderer, menu, and HID files and older recorded runs outside Sprint were outside the review. No local review checklist or sibling Tooling checkout was present. The supplied comments and reviews were empty.; nothing run (a code review).

Checked and fine:

  • Option ID 98 is unique in the checkout. The MkwVkHost and MkwVkRenderer layouts are unchanged, so the repository’s ABI-bump rule does not apply.
  • The existing Sprint row stays in its module. The new cell uses the existing grid, and Sprint remains eligible for both gamepad slots and keyboard bindings.
  • The player guide documents toggle_sprint=1. This binding-grid key is outside the module-row key registry.
  • The sole commit carries an AI co-author trailer. The diff contains source, documentation, and authored test cases with no game-derived files, third-party edits, build output, or generated shader edits.
  • The new isolated Sprint fixture honors CC and only compiles and runs its temporary authored fixture. The diff adds no downloads or test actions outside the fixture’s temporary files.
  • The guide manifest already differs from the base for the touched menu sources. The new cell still warrants a screenshot source comparison.

Testing: Read-only review. The author reports a Windows 11 llvm-mingw build and targeted test passes, with test_pads_after_menu failing on main too. There is no game run or Linux build result on this head. The supplied CI snapshot has six pending checks and no reported failures.

Overlaps and merge order: #1948 is already merged into the base. No pending overlap was identified from the checkout or supplied PR context.


Automated code review by BeanBot (gpt-6.2-sol), on the pull request's opening. Nothing was run; a maintainer decides.

**Recommendation: Request changes.** Sprint can latch again when focus returns, carry a note-button press into gameplay, or inject Sprint into an exclusive test run. The renderer also shares the new setting with the game thread without synchronization, and three existing test fixtures need updates. Adds a saved Toggle cell to Sprint’s Controls-page binding. Off preserves hold-to-sprint, and On lets players take their finger off the key or stick click while running. The UI follows the existing binding grid, but the new input latch needs to respect menu, focus, note, and automation boundaries. **`@fix` will do (16)** 1. **L1** `gpu/vk/vk_replay.cpp:7739`: The game thread reads `opts.sprint_toggle` through `rx_option` while the present thread can write it from Controls. (more on the line) 2. **L2** `host/nn_hid.c:1472`: During Insert or focus loss, `read_input` hides the held Sprint key, so the reset records `was = 0`. (more on the line) 3. **L3** `host/nn_hid.c:1474`: The note-wait check happens after the latch has already advanced. (more on the line) 4. **L4** `host/nn_hid.c:1431`: Script and bridge `SPRINT` now feed the player’s latch. A timed `SPRINT+LUP` followed by plain `LUP` keeps Sprint on after the timed press ends, contrary to the documented held input. An exclusive bridge movement can also inherit a latch set by a physical press before the request. Keep synthetic Sprint as a timed hold and suppress the player latch’s output while physical input is excluded. 5. **L5** (minor) `host/nn_hid.c:1468`: The latch only observes Sprint being disabled when a poll sees `sprint_enabled` false. (more on the line) 6. **L6** `tests/test_menu_keys.py:343`: The compiled fixture extracts the new load block without `has_toggle` or `toggle_set`, so compilation fails. Its checks at lines 356 and 358 also require the old inline predicates. Include the real helpers and keep coverage that Plus and Minus cannot be loaded, drawn, or selected as toggleable. 7. **L7** `tests/test_crt_options.py:370`: The compiled CRT fixture extracts `save` and the options loader without their new `toggle_on`, `has_toggle`, and `toggle_set` dependencies. The fixture fails to compile even though it only exercises CRT settings. Include the toggle helpers in its extracted code. 8. **L8** `tests/test_menu_save_atomic.py:386`: The Linux atomic-save fixture extracts `save()` without `toggle_on`. Its generated C++ fails to compile before exercising any disk-failure case. Include the getter and its Sprint setting in the fixture. 9. **L9** `tests/test_sprint.py:491`: The new wiring test matches source text, so removing the reset assignment or breaking the loader still passes. (more on the line) 10. **L10** (minor) `tests/test_walk_key.py:336`: The updated Walk fixture still ignores `CC` and can skip on a machine whose configured compiler has a different executable name or lives outside PATH. Reuse the Sprint fixture’s compiler selection. Apply the same selection to the default for the updated input-background fixture, whose suite declaration currently forces `gcc`. 11. **L11** (minor) `gpu/vk/vk_menu.cpp:3709`: The new saved key is accepted through `atoi`. (more on the line) 12. **L12** (minor) `docs/user/graphics-and-controls.md:266`: The main Sprint instructions and settings-table entry still say to hold the input and release it to stop. The Controls-tab Sprint help and the binding-cell help make the same unconditional claim. A player reading those after enabling Toggle gets the wrong instructions. Qualify the hold behavior as the default, point to the Toggle cell, and explain that the parent Sprint setting must be On. 13. **L13** (minor) `gpu/vk/vk_menu.cpp:2559`: The new whole-string help has no matching translation, so a player using a translated Controls page gets the Sprint Toggle instructions in English. (more on the line) 14. **L14** (minor) `gpu/vk/vk_menu.cpp:1396`: `HOS_KEY_SPRINT` is 31, which is the sign-bit position of a 32-bit `int`. (more on the line) 15. **L15** (minor): The PR ticks `mise run test` but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell. 16. **L17** (its part an edit can do) `host/nn_hid.c:1468`: Publish the toggle mode with the existing Sprint input snapshot and avoid a separate renderer option dispatch on each poll. **For the author (2)** (a run, a measurement or a decision) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1468`: The Measurements section calls this one extra compare, but with Sprint enabled and Toggle Off each poll also calls the renderer option dispatcher and updates edge state. (more on the line) **Checks:** Still running: `PR description / Description follows the template (pull_request_target)`, `CI / Conventional Commit title (pull_request)`, `CI / Workflow lint (actionlint) (pull_request)`, `CI / Website build and checks (pull_request)`, `Guide screenshots / Guide images and local UI source review (pull_request)`, `Third-party parts / As their notes say (pull_request)`. <details> <summary>More: does it help, other systems, switches, what was read</summary> **Does it help:** Unproven in game. The author reports a Windows 11 llvm-mingw build and passing targeted fixtures. The isolated latch test covers press, release, and reset with an input still visible to the helper. There is no physical-input run on this head, Linux build or start, or frame-time comparison against the base. The #1929 measurements predate this toggle. **Other systems:** Windows and Linux share the changed Npad polling path, including Steam Deck controller input. The pad and keyboard binding formats and existing Sprint A-press handoff are preserved in the diff. The new latch affects ocarina input and scripted or bridge-driven runs as detailed below. The renderer change is confined to the menu, settings and option callback. No render pass, save-game format, audio path, GPU capability check, or build configuration changes. **Switches:** No environment switches added or changed. The player setting `toggle_sprint` defaults to 0 (Off), which preserves held Sprint for existing settings files. `HOS_OPT_SPRINT_TOGGLE` is the host option ID for that setting. **Reviewed:** f73c817a5b58 against 6e1521a70a32: Read AGENTS.md, CONTRIBUTING.md, the Forgejo template, developer standards, the sole commit, the full eleven-file diff twice, and the surrounding code at each change. Followed the input and focus callers, binding snapshot, menu save/load and rollback, renderer option callback, and dependent test fixtures. Skimmed the guide manifest and translation tables. Unrelated portions of the large renderer, menu, and HID files and older recorded runs outside Sprint were outside the review. No local review checklist or sibling Tooling checkout was present. The supplied comments and reviews were empty.; nothing run (a code review). **Checked and fine:** - Option ID 98 is unique in the checkout. The `MkwVkHost` and `MkwVkRenderer` layouts are unchanged, so the repository’s ABI-bump rule does not apply. - The existing Sprint row stays in its module. The new cell uses the existing grid, and Sprint remains eligible for both gamepad slots and keyboard bindings. - The player guide documents `toggle_sprint=1`. This binding-grid key is outside the module-row key registry. - The sole commit carries an AI co-author trailer. The diff contains source, documentation, and authored test cases with no game-derived files, third-party edits, build output, or generated shader edits. - The new isolated Sprint fixture honors `CC` and only compiles and runs its temporary authored fixture. The diff adds no downloads or test actions outside the fixture’s temporary files. - The guide manifest already differs from the base for the touched menu sources. The new cell still warrants a screenshot source comparison. **Testing:** Read-only review. The author reports a Windows 11 llvm-mingw build and targeted test passes, with `test_pads_after_menu` failing on main too. There is no game run or Linux build result on this head. The supplied CI snapshot has six pending checks and no reported failures. **Overlaps and merge order:** #1948 is already merged into the base. No pending overlap was identified from the checkout or supplied PR context. </details> --- <sub>Automated code review by BeanBot (gpt-6.2-sol), on the pull request's opening. Nothing was run; a maintainer decides.</sub>
@ -1392,2 +1393,4 @@
controls[i].button == HOS_KEY_WALK);
}
// (the controls with a Toggle column: the Switch's buttons but Plus and Minus (opts.hold_toggle, a bit each), and Sprint
// (opts.sprint_toggle: its index is past an int's bits))
Member

L14 (minor): HOS_KEY_SPRINT is 31, which is the sign-bit position of a 32-bit int. Update the comment and PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids using its sign bit.

**L14** (minor): `HOS_KEY_SPRINT` is 31, which is the sign-bit position of a 32-bit `int`. Update the comment and PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids using its sign bit.
@ -2545,2 +2556,4 @@
const char* help = listening ? (listening == 1 ? "Press the gamepad button to use. Esc cancels." : "Press the key to use. Esc cancels.")
: page_row ? page_row->help
: sel_i < kControls && controls[sel_i].button == HOS_KEY_SPRINT && ccol == kCtlToggle
? "Toggle for Sprint: On, a press of its key or button starts Sprint and the next ends it; Off, Link "
Member

L13 (minor): The new whole-string help has no matching translation, so a player using a translated Controls page gets the Sprint Toggle instructions in English. Add the help to the menu translation tables once its wording is settled.

**L13** (minor): The new whole-string help has no matching translation, so a player using a translated Controls page gets the Sprint Toggle instructions in English. Add the help to the menu translation tables once its wording is settled.
@ -3696,2 +3708,2 @@
opts.hold_toggle = atoi(eq + 1) == 1 ? opts.hold_toggle | bit : opts.hold_toggle & ~bit;
}
for (int i = 0; i < HOS_KEY_N; i++)
if (has_toggle(i) && !strcmp(line + 7, bind_names[i])) toggle_set(i, atoi(eq + 1) == 1);
Member

L11 (minor): The new saved key is accepted through atoi. An out-of-range value such as toggle_sprint=4294967297 has undefined conversion behavior and can be mistaken for 1. Parse the numeric toggle value with a range-checked conversion and treat invalid values as Off.

**L11** (minor): The new saved key is accepted through `atoi`. An out-of-range value such as `toggle_sprint=4294967297` has undefined conversion behavior and can be mistaken for 1. Parse the numeric toggle value with a range-checked conversion and treat invalid values as Off.
@ -7736,6 +7736,7 @@ void fill_replay_table(MkwVkRenderer& t) {
if (id == HOS_OPT_MUTE_BACKGROUND) return opts.mute_background;
if (id == HOS_OPT_DISABLE_CAMERA_CENTERING) return opts.disable_camera_centering;
if (id == HOS_OPT_SPRINT) return opts.sprint;
if (id == HOS_OPT_SPRINT_TOGGLE) return opts.sprint_toggle;
Member

L1: The game thread reads opts.sprint_toggle through rx_option while the present thread can write it from Controls. The game-side read takes neither menu_mu nor another lock, so changing or resetting the Toggle while Npad polls is a C++ data race. The bridge’s whole-Options rollback also writes the field. Publish the host-facing value separately through an atomic or use one synchronization scheme for every read and write, including rollback.

**L1**: The game thread reads `opts.sprint_toggle` through `rx_option` while the present thread can write it from Controls. The game-side read takes neither `menu_mu` nor another lock, so changing or resetting the Toggle while Npad polls is a C++ data race. The bridge’s whole-`Options` rollback also writes the field. Publish the host-facing value separately through an atomic or use one synchronization scheme for every read and write, including rollback.
host/nn_hid.c Outdated
@ -1455,2 +1471,4 @@
!pads_in_background(getenv("HOS_INPUT_BACKGROUND"), rx_option(4, 0)));
const int sprint_held = sprint_toggle_step(sprint_enabled ? sprint_now : 0, sprint_toggle, sprint_reset);
/* (the ocarina's note wait, as oc_notes left it this read: Sprint lets go) */
sprint_step(&buttons, sprint_held, steered, menu_open || hos_oc_wait == 1, hos_now_ns());
Member

L3: The note-wait check happens after the latch has already advanced. A pad button shared by Sprint and a note can latch Sprint during the song and then start it when Link moves afterward, even if the button was released during the song. A shared keyboard key can also look newly pressed when note consumption ends while it is still held. Suppress latch changes during the note wait and require release of inputs carried out of it.

**L3**: The note-wait check happens after the latch has already advanced. A pad button shared by Sprint and a note can latch Sprint during the song and then start it when Link moves afterward, even if the button was released during the song. A shared keyboard key can also look newly pressed when note consumption ends while it is still held. Suppress latch changes during the note wait and require release of inputs carried out of it.
host/nn_hid.c Outdated
@ -1454,1 +1466,3 @@
const int sprint_held = sprint_enabled ? sprint_now : 0;
/* Sprint's Toggle (the Controls page): a press starts it and the next ends it, ended as the buttons' toggles are
(opening Insert, Plus or Minus, the focus lost) */
const int sprint_toggle = sprint_enabled && rx_option(HOS_OPT_SPRINT_TOGGLE, 0) == 1;
Member

L5 (minor): The latch only observes Sprint being disabled when a poll sees sprint_enabled false. With custom bindings, the bridge can set the Sprint row Off and back On between polls without any remap calling hos_input_clear_toggles, so a previous latch survives. Invalidate the latch when the host publishes a disable, even when the layout has no slots to move.

**L5** (minor): The latch only observes Sprint being disabled when a poll sees `sprint_enabled` false. With custom bindings, the bridge can set the Sprint row Off and back On between polls without any remap calling `hos_input_clear_toggles`, so a previous latch survives. Invalidate the latch when the host publishes a disable, even when the layout has no slots to move.
Member

L17 (minor): The Measurements section calls this one extra compare, but with Sprint enabled and Toggle Off each poll also calls the renderer option dispatcher and updates edge state. The standards ask for frame-time or FPS evidence with the feature on and off. Compare this head in both modes with the base in the same scene and build conditions, with repeated runs.

**L17** (minor): The Measurements section calls this one extra compare, but with Sprint enabled and Toggle Off each poll also calls the renderer option dispatcher and updates edge state. The standards ask for frame-time or FPS evidence with the feature on and off. Compare this head in both modes with the base in the same scene and build conditions, with repeated runs.
host/nn_hid.c Outdated
@ -1455,0 +1469,4 @@
const int sprint_reset = reset || pause != 0 ||
(sprint_toggle && !game_in_front() &&
!pads_in_background(getenv("HOS_INPUT_BACKGROUND"), rx_option(4, 0)));
const int sprint_held = sprint_toggle_step(sprint_enabled ? sprint_now : 0, sprint_toggle, sprint_reset);
Member

L2: During Insert or focus loss, read_input hides the held Sprint key, so the reset records was = 0. Hold Shift through an Insert open and close, or through an Alt-Tab away and back. The first visible sample sees a rising edge and restarts Sprint without a new press. Reuse the release-blocking behavior of HoldToggle or add equivalent rearming across suppressed input, including the first sample after focus or menu return.

**L2**: During Insert or focus loss, `read_input` hides the held Sprint key, so the reset records `was = 0`. Hold Shift through an Insert open and close, or through an Alt-Tab away and back. The first visible sample sees a rising edge and restarts Sprint without a new press. Reuse the release-blocking behavior of `HoldToggle` or add equivalent rearming across suppressed input, including the first sample after focus or menu return.
@ -443,0 +488,4 @@
self.assertEqual(build.returncode, 0, build.stderr)
run = subprocess.run([str(binary)], text=True, capture_output=True, timeout=30)
self.assertEqual(run.returncode, 0, run.stdout + run.stderr)
Member

L9: The new wiring test matches source text, so removing the reset assignment or breaking the loader still passes. The reset unit test supplies an uninterrupted held sample, and the Npad fixture keeps Sprint Toggle Off, so neither catches the actual focus or menu sequence. Exercise Toggle On through Npad with simulated input and menu/focus changes, and round-trip the real settings code through save, load, and Reset.

**L9**: The new wiring test matches source text, so removing the reset assignment or breaking the loader still passes. The reset unit test supplies an uninterrupted held sample, and the Npad fixture keeps Sprint Toggle Off, so neither catches the actual focus or menu sequence. Exercise Toggle On through Npad with simulated input and menu/focus changes, and round-trip the real settings code through save, load, and Reset.
Author
Member
@fix
Owner
@fix
Author
Member
@fix
Member

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is 38a9751023. It was already up to date with main.

Changed

  • Invalidate latch on disable: host/nn_hid.c:709 requests a toggle reset whenever Sprint is disabled, even if no bindings move.
  • Menu keys fixture: tests/test_menu_keys.py:245 extracts the toggle helpers and checks Plus and Minus are excluded from loading, drawing and selection.
  • CRT fixture dependencies: tests/test_crt_options.py:370 extracts the three toggle helpers.
  • Atomic-save fixture dependencies: tests/test_menu_save_atomic.py:71 includes the Sprint setting and extracts toggle_on.
  • Compiler selection: tests/test_walk_key.py:337 and tests/test_input_background.py:147 reuse Sprint's CC-aware compiler selection.
  • Range-safe toggle parsing: gpu/vk/vk_menu.cpp:3709 uses strtol and rejects invalid trailing text; tests/test_menu_keys.py:281 covers overflow and malformed values.
  • Sprint instructions: docs/user/graphics-and-controls.md:150, gpu/vk/menu/sprint.cpp:85 and gpu/vk/vk_menu.cpp:2563 explain the default hold behavior and Toggle cell.
  • Sprint toggle data race: Published the host-facing mode through the binding interface and synchronized rollback and Reset in gpu/vk/vk_menu.cpp:1401 and gpu/vk/vk_replay.cpp:7739.
  • Rearm after menu and focus transitions: Blocked the toggle until release after suppressed input in host/nn_hid.c:1013; added Npad transition coverage in tests/input_background_test.c:215.
  • Suppress toggle during note waits: Suspended latch changes during note waits and require release afterward in host/nn_hid.c:1484.
  • Synthetic Sprint holds: Kept synthetic Sprint separate from the physical latch and suppressed physical latch output during exclusive input in host/nn_hid.c:1443.
  • Sprint-specific reset: Disabling Sprint now schedules its own reset without clearing other button latches in host/nn_hid.c:712.
  • Missing compiler handling: Added C compiler preflight selection and a missing-compiler check in tests/run_all.py:137 and tests/test_input_background.py:147.

Changed in part (the rest is for the author)

  • Sprint mask comment and PR summary: gpu/vk/vk_menu.cpp:1396 corrects the sign-bit comment. The PR summary still needs the same correction.
  • Behavioral Sprint tests: Added Toggle On Npad menu and focus coverage in tests/input_background_test.c:215. Production save, reload and Reset still need behavioral coverage.
  • Sprint help translations: Added and checked the three French strings in gpu/vk/menu_i18n.h:1993 and tests/test_menu_i18n.py:203. The other language tables still need the three translations and corresponding coverage.
  • Sprint polling measurement: Published toggle mode with the input snapshot in host/nn_hid.c:1243. The author must compare repeated frame-time or FPS runs for base and head with Toggle On and Off in the same scene and build conditions.

Left for the author

  • Sprint toggle data race: Still needs synchronized publication of the host-facing value, including rollback.
  • Rearm after menu or focus loss: Still needs release blocking across suppressed input and the first returning sample.
  • Suppress toggle during note waits: Still needs latch suppression and release blocking for inputs carried out of a note wait.
  • Synthetic Sprint holds: Still needs synthetic Sprint separated from the player latch and latch output suppressed during exclusive input.
  • Sprint integration tests: Still needs Npad tests for menu and focus transitions and a real settings save, load and Reset round trip.
  • Translate Sprint Toggle help: Still needs translations for the settled help text in the menu language tables.
  • PR checklist and evidence: Update the PR description: clear the full-suite checkbox until it runs, record lint and named base failures, and restore the before-and-after screenshot item.
  • Sprint snapshot and performance measurement: Still needs toggle mode published with the Sprint input snapshot. Then compare repeated frame-time or FPS runs of base and updated head with Toggle on and off in the same scene and build conditions.
  • Guide screenshot source review: Review the affected UI, guide and fixture against the recorded image, then update the review records and hashes as appropriate. No screenshot review or hash update was performed.
  • PR summary: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. The PR description is outside the checkout.
  • PR checklist and evidence: Clear the full-suite checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description. The description is outside the checkout and no checks were run.
  • Guide screenshot check: Review the affected UI, guide images and changed fixture sources, then update the review records or recapture affected images as appropriate. The required visual review and capture were not run.

Still for the author (no edit does these)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.

Nothing was built or run: check the change. Its review follows.


Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is 38a97510239b. It was already up to date with `main`. **Changed** - **Invalidate latch on disable**: host/nn_hid.c:709 requests a toggle reset whenever Sprint is disabled, even if no bindings move. - **Menu keys fixture**: tests/test_menu_keys.py:245 extracts the toggle helpers and checks Plus and Minus are excluded from loading, drawing and selection. - **CRT fixture dependencies**: tests/test_crt_options.py:370 extracts the three toggle helpers. - **Atomic-save fixture dependencies**: tests/test_menu_save_atomic.py:71 includes the Sprint setting and extracts toggle_on. - **Compiler selection**: tests/test_walk_key.py:337 and tests/test_input_background.py:147 reuse Sprint's CC-aware compiler selection. - **Range-safe toggle parsing**: gpu/vk/vk_menu.cpp:3709 uses strtol and rejects invalid trailing text; tests/test_menu_keys.py:281 covers overflow and malformed values. - **Sprint instructions**: docs/user/graphics-and-controls.md:150, gpu/vk/menu/sprint.cpp:85 and gpu/vk/vk_menu.cpp:2563 explain the default hold behavior and Toggle cell. - **Sprint toggle data race**: Published the host-facing mode through the binding interface and synchronized rollback and Reset in gpu/vk/vk_menu.cpp:1401 and gpu/vk/vk_replay.cpp:7739. - **Rearm after menu and focus transitions**: Blocked the toggle until release after suppressed input in host/nn_hid.c:1013; added Npad transition coverage in tests/input_background_test.c:215. - **Suppress toggle during note waits**: Suspended latch changes during note waits and require release afterward in host/nn_hid.c:1484. - **Synthetic Sprint holds**: Kept synthetic Sprint separate from the physical latch and suppressed physical latch output during exclusive input in host/nn_hid.c:1443. - **Sprint-specific reset**: Disabling Sprint now schedules its own reset without clearing other button latches in host/nn_hid.c:712. - **Missing compiler handling**: Added C compiler preflight selection and a missing-compiler check in tests/run_all.py:137 and tests/test_input_background.py:147. **Changed in part (the rest is for the author)** - **Sprint mask comment and PR summary**: gpu/vk/vk_menu.cpp:1396 corrects the sign-bit comment. The PR summary still needs the same correction. - **Behavioral Sprint tests**: Added Toggle On Npad menu and focus coverage in tests/input_background_test.c:215. Production save, reload and Reset still need behavioral coverage. - **Sprint help translations**: Added and checked the three French strings in gpu/vk/menu_i18n.h:1993 and tests/test_menu_i18n.py:203. The other language tables still need the three translations and corresponding coverage. - **Sprint polling measurement**: Published toggle mode with the input snapshot in host/nn_hid.c:1243. The author must compare repeated frame-time or FPS runs for base and head with Toggle On and Off in the same scene and build conditions. **Left for the author** - **Sprint toggle data race**: Still needs synchronized publication of the host-facing value, including rollback. - **Rearm after menu or focus loss**: Still needs release blocking across suppressed input and the first returning sample. - **Suppress toggle during note waits**: Still needs latch suppression and release blocking for inputs carried out of a note wait. - **Synthetic Sprint holds**: Still needs synthetic Sprint separated from the player latch and latch output suppressed during exclusive input. - **Sprint integration tests**: Still needs Npad tests for menu and focus transitions and a real settings save, load and Reset round trip. - **Translate Sprint Toggle help**: Still needs translations for the settled help text in the menu language tables. - **PR checklist and evidence**: Update the PR description: clear the full-suite checkbox until it runs, record lint and named base failures, and restore the before-and-after screenshot item. - **Sprint snapshot and performance measurement**: Still needs toggle mode published with the Sprint input snapshot. Then compare repeated frame-time or FPS runs of base and updated head with Toggle on and off in the same scene and build conditions. - **Guide screenshot source review**: Review the affected UI, guide and fixture against the recorded image, then update the review records and hashes as appropriate. No screenshot review or hash update was performed. - **PR summary**: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. The PR description is outside the checkout. - **PR checklist and evidence**: Clear the full-suite checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description. The description is outside the checkout and no checks were run. - **Guide screenshot check**: Review the affected UI, guide images and changed fixture sources, then update the review records or recapture affected images as appropriate. The required visual review and capture were not run. **Still for the author** (no edit does these) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. Nothing was built or run: check the change. Its review follows. --- <sub>Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.</sub>
fix(controls): correct Sprint settings and fixture dependencies
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 8s
PR description / Description follows the template (pull_request_target) Successful in 11s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 12s
CI / Workflow lint (actionlint) (pull_request) Successful in 13s
Third-party parts / As their notes say (pull_request) Successful in 15s
CI / Website build and checks (pull_request) Has been cancelled
38a9751023
As the review on #1977 asked.
BeanBot requested changes 2026-11-21 02:45:57 +00:00
Dismissed
BeanBot left a comment

Recommendation: Request changes. A held Sprint key can still restart Sprint when focus returns during an input poll, and an exclusive bridge request can inherit the player's Sprint-to-A handoff. The new fixture covers steady focus transitions and timed synthetic Sprint, but misses both cases.

Adds a saved Sprint Toggle to Controls so players can start and stop Sprint with a press. Held mode remains the default. The follow-up removes the renderer option dispatch from Npad polling and adds input transition coverage, at the cost of more state around focus, menus and synthetic input.

@fix will do (9)

  1. L2 host/nn_hid.c:1486: Rearming uses a later focus check than the one that gated the keyboard sample. Hold Shift through Alt-Tab and regain focus after read_input takes the background path but before npad_states checks focus. The hidden zero sample clears blocked, and the next poll latches Sprint without a new press. Carry the sample's suppression state from read_input into the latch and rearm only after a readable release.
  2. L4 host/nn_hid.c:1489: Masking the physical latch leaves its A-button handoff state active. While moving with Sprint latched, start an exclusive bridge A press without SPRINT. sprint_step sees running from the physical Sprint and clears the bridge's A for the 50 ms roll gap. A pending handoff can also inject A into a movement-only request. Clear the previous source's handoff state when entering exclusive input, then process the current synthetic input normally.
  3. L9 tests/test_menu_keys.py:285: The loader test starts from hand-written settings, so it still passes if production save omits the key or Controls Reset leaves the previous value on disk or in the host. Exercise the real save and reload with Toggle On, then exercise production Reset and verify the reloaded and host-facing values are Off.
  4. L13 (minor) gpu/vk/menu_i18n.h:2474: The three new Sprint help strings are translated only in French. A player using Spanish, for example, sees the Toggle help in English. Add the complete strings to the remaining language tables and extend the translation check to each table.
  5. L14 (minor) gpu/vk/vk_menu.cpp: The code comment is fixed, but the PR summary still says Sprint's index is past the mask's bits. Index 31 is the sign-bit position of a 32-bit int. Update the summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit.
  6. L15 (minor): The PR ticks mise run test but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell.
  7. L17 (its part an edit can do) host/nn_hid.c:1482: Reuse the focus state from the input sample and consume the Sprint reset only when needed.
  8. L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.
  9. L21 (minor) tests/run_all.py:137: Suite preflight treats CC as a single executable, while the standalone fixture accepts a compiler command with flags. (more on the line)

For the author (2) (a run, a measurement or a decision)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1482: The dispatch is gone, but the polling path now takes a reset exchange even with Toggle Off and can query focus twice with Toggle On. Compare base and head with repeated interleaved runs in the same scene and build conditions. Report FPS and frame times for both head modes.

Since the last review (on f73c817a5b)
Resolved: L1 (The game thread reads opts.sprint_toggle through rx_option while the present thread ca…); L3 (The note-wait check happens after the latch has already advanced. A pad button shared by S…); L5 (The latch only observes Sprint being disabled when a poll sees sprint_enabled false. Wit…); L6 (The compiled fixture extracts the new load block without has_toggle or toggle_set, so…); L7 (The compiled CRT fixture extracts save and the options loader without their new toggle_…); L8 (The Linux atomic-save fixture extracts save()withouttoggle_on. Its generated C++ fai…); L10 (The updated Walk fixture still ignores CCand can skip on a machine whose configured com…); L11 (The new saved key is accepted throughatoi. An out-of-range value such as toggle_sprint…); L12 (The main Sprint instructions and settings-table entry still say to hold the input and rele…)
Still open: L2 (Rearming uses a later focus check than the one that gated the keyboard sample. Hold Shift…); L4 (Masking the physical latch leaves its A-button handoff state active. While moving with Spr…); L9 (The loader test starts from hand-written settings, so it still passes if production save o…); L13 (The three new Sprint help strings are translated only in French. A player using Spanish, f…); L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks mise run test but lists only targeted results, and the source shows additio…); L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The dispatch is gone, but the polling path now takes a reset exchange even with Toggle Off…)
New: L18 (The check Guide screenshots / Guide images and local UI source review (pull_request) fai…); L19 (Disabling Sprint now clears all button latches. With custom Sprint bindings and ZL latched…); L20 (With no CC, GCC or Clang available, compiler selection returns [None] and the fixture pa…); L21 (Suite preflight treats CC as a single executable, while the standalone fixture accepts a c…)

Checks: Still running: CI / Website build and checks (pull_request).

More: does it help, other systems, switches, what was read

Does it help: Unproven. The PR reports a Windows build and targeted game-free checks for the original commit. No physical-input game run or repeated, same-scene FPS or frame-time comparison was supplied for base 6e1521a7 and head 2ad82e2a with Toggle On and Off. The #1929 result predates this change.

Other systems: Windows and Linux share the Npad path, including Steam Deck and background controller input. The change also touches ocarina note waits, the control bridge, scripted input, settings persistence and the guide's recorded UI sources. Host and renderer must be updated together for API version 40. The diff leaves GPU rendering passes, audio and game save data alone.

Switches: toggle_sprint is a saved Controls preference, Off by default. Off preserves the previous held behavior for existing settings files. The Controls tab's Sprint enable setting still defaults to Off. The player guide documents the new preference. No environment switch was added or changed.

Reviewed: 38a9751023 against 6e1521a70a: Read AGENTS.md, CONTRIBUTING.md, the Forgejo PR template and the developer standards in full. No docs/review-checklist.md was present. Read the three PR commits and the changed sections of all 20 files, including the surrounding option publication, save/load, Reset, focus, note wait, bridge and fixture code. Skimmed unchanged translation tables, guide capture records and unrelated renderer code. Local Tooling rules were unavailable. The supplied PR body identifies f73c817a, while the checkout and range-diff end at 2ad82e2a. Ledger findings use the checkout head.; nothing run (a code review).

Checked and fine:

  • The versioned binding opcode is appended, and renderer calls guard it with API version 40.
  • The Sprint Toggle uses a separate field, keeps Plus and Minus out of the Toggle column, and leaves both gamepad slots and the keyboard binding editable.
  • The loader defaults missing or invalid Sprint Toggle values to Off.
  • The diff has no whitespace errors, generated files, third-party changes or new download steps.

Testing: Read-only review. The PR reports targeted llvm-mingw checks and a Windows build for the original commit, with test_pads_after_menu named as a base failure. Head lint, the full suite, Linux build and physical-device runs remain unverified. Supplied CI reports the guide screenshot source review failed and the website check is still running.

Overlaps and merge order: #1948 is already in the base. No other overlap was found in the supplied record.


Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.

**Recommendation: Request changes.** A held Sprint key can still restart Sprint when focus returns during an input poll, and an exclusive bridge request can inherit the player's Sprint-to-A handoff. The new fixture covers steady focus transitions and timed synthetic Sprint, but misses both cases. Adds a saved Sprint Toggle to Controls so players can start and stop Sprint with a press. Held mode remains the default. The follow-up removes the renderer option dispatch from Npad polling and adds input transition coverage, at the cost of more state around focus, menus and synthetic input. **`@fix` will do (9)** 1. **L2** `host/nn_hid.c:1486`: Rearming uses a later focus check than the one that gated the keyboard sample. Hold Shift through Alt-Tab and regain focus after read_input takes the background path but before npad_states checks focus. The hidden zero sample clears blocked, and the next poll latches Sprint without a new press. Carry the sample's suppression state from read_input into the latch and rearm only after a readable release. 2. **L4** `host/nn_hid.c:1489`: Masking the physical latch leaves its A-button handoff state active. While moving with Sprint latched, start an exclusive bridge A press without SPRINT. sprint_step sees running from the physical Sprint and clears the bridge's A for the 50 ms roll gap. A pending handoff can also inject A into a movement-only request. Clear the previous source's handoff state when entering exclusive input, then process the current synthetic input normally. 3. **L9** `tests/test_menu_keys.py:285`: The loader test starts from hand-written settings, so it still passes if production save omits the key or Controls Reset leaves the previous value on disk or in the host. Exercise the real save and reload with Toggle On, then exercise production Reset and verify the reloaded and host-facing values are Off. 4. **L13** (minor) `gpu/vk/menu_i18n.h:2474`: The three new Sprint help strings are translated only in French. A player using Spanish, for example, sees the Toggle help in English. Add the complete strings to the remaining language tables and extend the translation check to each table. 5. **L14** (minor) `gpu/vk/vk_menu.cpp`: The code comment is fixed, but the PR summary still says Sprint's index is past the mask's bits. Index 31 is the sign-bit position of a 32-bit int. Update the summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. 6. **L15** (minor): The PR ticks `mise run test` but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell. 7. **L17** (its part an edit can do) `host/nn_hid.c:1482`: Reuse the focus state from the input sample and consume the Sprint reset only when needed. 8. **L18**: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s. 9. **L21** (minor) `tests/run_all.py:137`: Suite preflight treats CC as a single executable, while the standalone fixture accepts a compiler command with flags. (more on the line) **For the author (2)** (a run, a measurement or a decision) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1482`: The dispatch is gone, but the polling path now takes a reset exchange even with Toggle Off and can query focus twice with Toggle On. Compare base and head with repeated interleaved runs in the same scene and build conditions. Report FPS and frame times for both head modes. **Since the last review** (on f73c817a5b58) Resolved: L1 (The game thread reads `opts.sprint_toggle` through `rx_option` while the present thread ca…); L3 (The note-wait check happens after the latch has already advanced. A pad button shared by S…); L5 (The latch only observes Sprint being disabled when a poll sees `sprint_enabled` false. Wit…); L6 (The compiled fixture extracts the new load block without `has_toggle` or `toggle_set`, so…); L7 (The compiled CRT fixture extracts `save` and the options loader without their new `toggle_…); L8 (The Linux atomic-save fixture extracts `save()` without `toggle_on`. Its generated C++ fai…); L10 (The updated Walk fixture still ignores `CC` and can skip on a machine whose configured com…); L11 (The new saved key is accepted through `atoi`. An out-of-range value such as `toggle_sprint…); L12 (The main Sprint instructions and settings-table entry still say to hold the input and rele…) Still open: L2 (Rearming uses a later focus check than the one that gated the keyboard sample. Hold Shift…); L4 (Masking the physical latch leaves its A-button handoff state active. While moving with Spr…); L9 (The loader test starts from hand-written settings, so it still passes if production save o…); L13 (The three new Sprint help strings are translated only in French. A player using Spanish, f…); L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks `mise run test` but lists only targeted results, and the source shows additio…); L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The dispatch is gone, but the polling path now takes a reset exchange even with Toggle Off…) New: L18 (The check `Guide screenshots / Guide images and local UI source review (pull_request)` fai…); L19 (Disabling Sprint now clears all button latches. With custom Sprint bindings and ZL latched…); L20 (With no CC, GCC or Clang available, compiler selection returns `[None]` and the fixture pa…); L21 (Suite preflight treats CC as a single executable, while the standalone fixture accepts a c…) **Checks:** Still running: `CI / Website build and checks (pull_request)`. <details> <summary>More: does it help, other systems, switches, what was read</summary> **Does it help:** Unproven. The PR reports a Windows build and targeted game-free checks for the original commit. No physical-input game run or repeated, same-scene FPS or frame-time comparison was supplied for base 6e1521a7 and head 2ad82e2a with Toggle On and Off. The #1929 result predates this change. **Other systems:** Windows and Linux share the Npad path, including Steam Deck and background controller input. The change also touches ocarina note waits, the control bridge, scripted input, settings persistence and the guide's recorded UI sources. Host and renderer must be updated together for API version 40. The diff leaves GPU rendering passes, audio and game save data alone. **Switches:** `toggle_sprint` is a saved Controls preference, Off by default. Off preserves the previous held behavior for existing settings files. The Controls tab's Sprint enable setting still defaults to Off. The player guide documents the new preference. No environment switch was added or changed. **Reviewed:** 38a97510239b against 6e1521a70a32: Read AGENTS.md, CONTRIBUTING.md, the Forgejo PR template and the developer standards in full. No docs/review-checklist.md was present. Read the three PR commits and the changed sections of all 20 files, including the surrounding option publication, save/load, Reset, focus, note wait, bridge and fixture code. Skimmed unchanged translation tables, guide capture records and unrelated renderer code. Local Tooling rules were unavailable. The supplied PR body identifies f73c817a, while the checkout and range-diff end at 2ad82e2a. Ledger findings use the checkout head.; nothing run (a code review). **Checked and fine:** - The versioned binding opcode is appended, and renderer calls guard it with API version 40. - The Sprint Toggle uses a separate field, keeps Plus and Minus out of the Toggle column, and leaves both gamepad slots and the keyboard binding editable. - The loader defaults missing or invalid Sprint Toggle values to Off. - The diff has no whitespace errors, generated files, third-party changes or new download steps. **Testing:** Read-only review. The PR reports targeted llvm-mingw checks and a Windows build for the original commit, with test_pads_after_menu named as a base failure. Head lint, the full suite, Linux build and physical-device runs remain unverified. Supplied CI reports the guide screenshot source review failed and the website check is still running. **Overlaps and merge order:** #1948 is already in the base. No other overlap was found in the supplied record. </details> --- <sub>Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.</sub>
@ -708,6 +709,11 @@ static void bind_one(int set, int button, int slot, int code) {
}
if (set == HOS_BIND_SPRINT_ON) {
InterlockedExchange(&sprint_on, code != 0);
Member

L19 (minor): Disabling Sprint now clears all button latches. With custom Sprint bindings and ZL latched, turning Sprint Off through the bridge drops Link's shield even when no binding moves and the menu stays closed. Use a Sprint-specific pending reset and leave the shared reset for events that should release all buttons.

**L19** (minor): Disabling Sprint now clears all button latches. With custom Sprint bindings and ZL latched, turning Sprint Off through the bridge drops Link's shield even when no binding moves and the menu stays closed. Use a Sprint-specific pending reset and leave the shared reset for events that should release all buttons.
tests/run_all.py Outdated
@ -134,6 +134,7 @@ def find_tools():
"spirv_val": which("spirv-val"),
"glslang": which("glslangValidator", "glslang"),
"dotnet": which("dotnet"),
"cc": which(os.environ.get("CC", ""), "gcc", "clang"),
Member

L21 (minor): Suite preflight treats CC as a single executable, while the standalone fixture accepts a compiler command with flags. With CC='clang -m64' and both GCC and Clang installed, the suite silently selects GCC and drops the requested flags. Use the same compiler command parsing for preflight and execution.

**L21** (minor): Suite preflight treats CC as a single executable, while the standalone fixture accepts a compiler command with flags. With CC='clang -m64' and both GCC and Clang installed, the suite silently selects GCC and drops the requested flags. Use the same compiler command parsing for preflight and execution.
@ -142,3 +145,3 @@
exe = folder / "fixture.exe"
# Production's button table deliberately leaves its unused axis fields zero-initialized.
run([a.cc, "-std=c11", "-include", ENV_ON_H, "-Wall", "-Wextra", "-Werror", "-Wno-missing-field-initializers",
cc = [a.cc] if a.cc else compiler_command("CC", "gcc", "clang")
Member

L20 (minor): With no CC, GCC or Clang available, compiler selection returns [None] and the fixture passes it to subprocess. The suite previously reported the missing compiler during preflight. Check the selection before invoking subprocess and report the missing dependency through the suite's check handling.

**L20** (minor): With no CC, GCC or Clang available, compiler selection returns `[None]` and the fixture passes it to subprocess. The suite previously reported the missing compiler during preflight. Check the selection before invoking subprocess and report the missing dependency through the suite's check handling.
Author
Member
@fix
Member

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is 2079d6469f. It was already up to date with main.

Changed

  • Sprint rearming after focus loss: host/nn_hid.c:1286 carries the sampled focus state into the latch; tests/input_background_test.c:232 covers focus returning after a suppressed sample.
  • Exclusive input handoff: host/nn_hid.c:1491 clears the previous Sprint A handoff on entry to exclusive input before processing synthetic input; the input fixture covers both an A press and a pending handoff.
  • Compiler command preflight: tests/run_all.py:139 and tests/compiler_command.py share compiler-command parsing with the fixture, retaining CC flags; a regression check was added. No tests or builds were run.
  • Sprint A handoff, host/nn_hid.c:1035: Preserved the last Sprint-supplied A state while clearing a pending tap, and added a packet and press-edge check in tests/input_background_test.c:237.
  • Exclusive input and focus, host/nn_hid.c:1485: Distinguished foreground exclusion from focus loss, suppressed physical Sprint during exclusion and added fixture coverage for latch retention and actual focus loss.
  • Compiler preflight, tests/run_all.py:268: Report malformed compiler quoting through preflight and added a regression fixture. No tests were run.
  • Trailing blank line, tests/compiler_command.py:57: Removed the trailing blank line.

Changed in part (the rest is for the author)

  • Polling overhead measurement: host/nn_hid.c:1485 reuses sampled focus and exchanges the Sprint reset only when Toggle is enabled. The author must compare base and head with repeated interleaved runs in the same scene and build conditions, reporting FPS and frame times for both head modes.
  • Held-mode bookkeeping, host/nn_hid.c:1485: Skipped the latch helper in steady held mode while retaining transition handling. The author must compare base and both head modes in repeated interleaved runs on the same hardware, scene, settings and build type, and report average FPS, 1% lows and frame times.

Left for the author

  • Save and Reset round trip: The production save, reload and Controls Reset round-trip test remains to be written in tests/test_menu_keys.py; verify Toggle On survives reload and Reset saves and publishes Off.
  • Sprint help translations: The three complete Sprint help strings still need translations in the twelve remaining language tables, and the translation check must cover every table.
  • PR summary wording: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit; the summary is outside this checkout.
  • PR validation and screenshots: Clear the full-suite checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description; the description is outside this checkout.
  • Guide screenshot review check: Review the changed UI and fixture sources against the Graphics guide image, then update the review record and hashes or recapture the image. The recorded review revision is unavailable in this checkout, so its broad source drift was not approved.
  • Save and reset round trip, tests/test_menu_keys.py:285: The production save and Reset round-trip fixture remains unwritten; add it and verify the reloaded and host-facing values.
  • Sprint help translations, gpu/vk/menu_i18n.h:2474: The remaining language translations and per-table translation checks remain unwritten.
  • PR summary: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit; PR metadata is outside the checkout.
  • PR checklist: Clear the full-suite checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description; PR metadata is outside the checkout.
  • Guide screenshot review: Review the affected UI and guide images, then update the review fingerprints for the affected sources and fixtures. The check requires that review before hashes are updated.

Still for the author (no edit does these)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Nothing was built or run: check the change. Its review follows.


Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is 2079d6469f50. It was already up to date with `main`. **Changed** - **Sprint rearming after focus loss**: host/nn_hid.c:1286 carries the sampled focus state into the latch; tests/input_background_test.c:232 covers focus returning after a suppressed sample. - **Exclusive input handoff**: host/nn_hid.c:1491 clears the previous Sprint A handoff on entry to exclusive input before processing synthetic input; the input fixture covers both an A press and a pending handoff. - **Compiler command preflight**: tests/run_all.py:139 and tests/compiler_command.py share compiler-command parsing with the fixture, retaining CC flags; a regression check was added. No tests or builds were run. - **Sprint A handoff, host/nn_hid.c:1035**: Preserved the last Sprint-supplied A state while clearing a pending tap, and added a packet and press-edge check in tests/input_background_test.c:237. - **Exclusive input and focus, host/nn_hid.c:1485**: Distinguished foreground exclusion from focus loss, suppressed physical Sprint during exclusion and added fixture coverage for latch retention and actual focus loss. - **Compiler preflight, tests/run_all.py:268**: Report malformed compiler quoting through preflight and added a regression fixture. No tests were run. - **Trailing blank line, tests/compiler_command.py:57**: Removed the trailing blank line. **Changed in part (the rest is for the author)** - **Polling overhead measurement**: host/nn_hid.c:1485 reuses sampled focus and exchanges the Sprint reset only when Toggle is enabled. The author must compare base and head with repeated interleaved runs in the same scene and build conditions, reporting FPS and frame times for both head modes. - **Held-mode bookkeeping, host/nn_hid.c:1485**: Skipped the latch helper in steady held mode while retaining transition handling. The author must compare base and both head modes in repeated interleaved runs on the same hardware, scene, settings and build type, and report average FPS, 1% lows and frame times. **Left for the author** - **Save and Reset round trip**: The production save, reload and Controls Reset round-trip test remains to be written in tests/test_menu_keys.py; verify Toggle On survives reload and Reset saves and publishes Off. - **Sprint help translations**: The three complete Sprint help strings still need translations in the twelve remaining language tables, and the translation check must cover every table. - **PR summary wording**: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit; the summary is outside this checkout. - **PR validation and screenshots**: Clear the full-suite checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description; the description is outside this checkout. - **Guide screenshot review check**: Review the changed UI and fixture sources against the Graphics guide image, then update the review record and hashes or recapture the image. The recorded review revision is unavailable in this checkout, so its broad source drift was not approved. - **Save and reset round trip, tests/test_menu_keys.py:285**: The production save and Reset round-trip fixture remains unwritten; add it and verify the reloaded and host-facing values. - **Sprint help translations, gpu/vk/menu_i18n.h:2474**: The remaining language translations and per-table translation checks remain unwritten. - **PR summary**: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit; PR metadata is outside the checkout. - **PR checklist**: Clear the full-suite checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description; PR metadata is outside the checkout. - **Guide screenshot review**: Review the affected UI and guide images, then update the review fingerprints for the affected sources and fixtures. The check requires that review before hashes are updated. **Still for the author** (no edit does these) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. Nothing was built or run: check the change. Its review follows. --- <sub>Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.</sub>
fix(input): preserve Sprint input state across focus and exclusive changes
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 8s
PR description / Description follows the template (pull_request_target) Successful in 12s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 11s
CI / Workflow lint (actionlint) (pull_request) Successful in 12s
Third-party parts / As their notes say (pull_request) Successful in 13s
CI / Website build and checks (pull_request) Has been cancelled
2079d6469f
As the review on #1977 asked.
BeanBot requested changes 2026-11-21 02:57:24 +00:00
Dismissed
BeanBot left a comment

Recommendation: Request changes. A second exclusive takeover can still erase a bridge A tap before the game sees it. The updated input fixture also expects a Sprint press to count before the key has been rearmed.

The Controls page lets players choose press-to-toggle Sprint while keeping held Sprint as the default. HEAD preserves the physical latch during foreground exclusive input, adds an A release and replay, and skips latch bookkeeping in steady held mode. The extra handoff state still needs to distinguish physical taps from bridge taps.

@fix will do (7)

  1. L4 host/nn_hid.c:1498: Start a one-frame exclusive bridge A tap immediately after Sprint supplied A. The first packet correctly releases A. If a second exclusive movement request starts after one nonexclusive sample and before the 50 ms gap expires, it clears the queued bridge tap and the game never sees A pressed. Track the source of the pending tap so takeover discards stale physical input and preserves an accepted bridge replay. Add a fixture case that drops bridge A during the gap and checks the press edge across the second takeover.
  2. L9 tests/test_menu_keys.py:285: The loader test starts from hand-written settings, so it still passes if production save omits the key or Controls Reset leaves the previous value on disk or in the host. Exercise the real save and reload with Toggle On, then exercise production Reset and verify the reloaded and host-facing values are Off.
  3. L13 (minor) gpu/vk/menu_i18n.h:2474: The three new Sprint help strings are translated only in French. A player using Spanish, for example, sees the Toggle help in English. Add the complete strings to the remaining language tables and extend the translation check to each table.
  4. L14 (minor) gpu/vk/vk_menu.cpp: The code comment is fixed, but the PR summary still says Sprint's index is past the mask's bits. Index 31 is the sign-bit position of a 32-bit int. Update the summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit.
  5. L15 (minor): The PR ticks mise run test but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell.
  6. L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.
  7. L25 tests/input_background_test.c:254: The fixture ends exclusion and presses G on the same sample, then expects Sprint to turn off. (more on the line)

For the author (2) (a run, a measurement or a decision)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Since the last review (on 38a9751023)
Resolved: L2 (Rearming uses a later focus check than the one that gated the keyboard sample. Hold Shift…); L21 (Suite preflight treats CC as a single executable, while the standalone fixture accepts a c…)
Still open: L4 (Start a one-frame exclusive bridge A tap immediately after Sprint supplied A. The first pa…); L9 (The loader test starts from hand-written settings, so it still passes if production save o…); L13 (The three new Sprint help strings are translated only in French. A player using Spanish, f…); L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks mise run test but lists only targeted results, and the source shows additio…); L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…); L18 (The check Guide screenshots / Guide images and local UI source review (pull_request) fai…)
New: L22 (Both exclusive paths return before assigning input_in_front, so a skipped focus query now…); L23 (On Linux, a CC value with an unmatched quote raises ValueError in shlex.split during prefl…); L24 (The trailing blank line is flagged by git diff --check. Remove it.); L25 (The fixture ends exclusion and presses G on the same sample, then expects Sprint to turn o…)

Checks: Still running: CI / Website build and checks (pull_request).

More: does it help, other systems, switches, what was read

Does it help: Unproven on 9842f922. The PR reports an earlier Windows build and targeted checks, with no in-game run or frame timings for this head. The cited #1929 measurements cover the earlier Sprint implementation.

Other systems: The shared Npad path affects keyboard, controllers, Steam Deck, background input, and scripted and bridge input on Windows and Linux. The renderer API bump requires matching host and renderer builds. The diff touches no GPU pass, save format, or audio path. Physical-device and frame-pacing evidence is still missing.

Switches: Sprint's Controls-page Toggle is saved as toggle_sprint and defaults to Off, preserving held input for existing profiles. This is a player preference, with no new environment switch. The guide documents it.

Reviewed: 2079d6469f against 6e1521a70a: Reviewed checkout HEAD 9842f922 against ledger head 9482fb1760a5. Read the six changed files through their affected functions and fixtures, the full base-to-HEAD diff, repository rules and PR template, and the relevant menu save, load, Reset, focus, bridge and A-edge code. Skimmed translation tables and screenshot maintenance records. Unchanged portions of large files, screenshot pixels and external Tooling rules were outside the read.; nothing run (a code review).

Checked and fine:

  • The basic exclusive A handoff produces a released packet followed by an unmasked A press. A stale physical tap is cleared on a movement-only takeover.
  • Existing INI files default to held Sprint. Production Reset clears the renderer option and publishes Off to the host.
  • The renderer API version is bumped to 40, and renderer calls to the new binding are version-gated.
  • No third-party, generated, shader, game-data, or build-output changes appear in the diff.

Testing: Read-only review. Both git diff --check comparisons passed. The PR reports earlier Windows llvm-mingw and targeted check results. No builds, tests or game runs were performed here. The supplied CI lists the guide screenshot check as failing and the website check as running.

Overlaps and merge order: The prerequisite #1948 is merged at the base. No other overlap found in the checkout or supplied review data.


Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.

**Recommendation: Request changes.** A second exclusive takeover can still erase a bridge A tap before the game sees it. The updated input fixture also expects a Sprint press to count before the key has been rearmed. The Controls page lets players choose press-to-toggle Sprint while keeping held Sprint as the default. HEAD preserves the physical latch during foreground exclusive input, adds an A release and replay, and skips latch bookkeeping in steady held mode. The extra handoff state still needs to distinguish physical taps from bridge taps. **`@fix` will do (7)** 1. **L4** `host/nn_hid.c:1498`: Start a one-frame exclusive bridge A tap immediately after Sprint supplied A. The first packet correctly releases A. If a second exclusive movement request starts after one nonexclusive sample and before the 50 ms gap expires, it clears the queued bridge tap and the game never sees A pressed. Track the source of the pending tap so takeover discards stale physical input and preserves an accepted bridge replay. Add a fixture case that drops bridge A during the gap and checks the press edge across the second takeover. 2. **L9** `tests/test_menu_keys.py:285`: The loader test starts from hand-written settings, so it still passes if production save omits the key or Controls Reset leaves the previous value on disk or in the host. Exercise the real save and reload with Toggle On, then exercise production Reset and verify the reloaded and host-facing values are Off. 3. **L13** (minor) `gpu/vk/menu_i18n.h:2474`: The three new Sprint help strings are translated only in French. A player using Spanish, for example, sees the Toggle help in English. Add the complete strings to the remaining language tables and extend the translation check to each table. 4. **L14** (minor) `gpu/vk/vk_menu.cpp`: The code comment is fixed, but the PR summary still says Sprint's index is past the mask's bits. Index 31 is the sign-bit position of a 32-bit int. Update the summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. 5. **L15** (minor): The PR ticks `mise run test` but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell. 6. **L18**: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s. 7. **L25** `tests/input_background_test.c:254`: The fixture ends exclusion and presses G on the same sample, then expects Sprint to turn off. (more on the line) **For the author (2)** (a run, a measurement or a decision) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. **Since the last review** (on 38a97510239b) Resolved: L2 (Rearming uses a later focus check than the one that gated the keyboard sample. Hold Shift…); L21 (Suite preflight treats CC as a single executable, while the standalone fixture accepts a c…) Still open: L4 (Start a one-frame exclusive bridge A tap immediately after Sprint supplied A. The first pa…); L9 (The loader test starts from hand-written settings, so it still passes if production save o…); L13 (The three new Sprint help strings are translated only in French. A player using Spanish, f…); L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks `mise run test` but lists only targeted results, and the source shows additio…); L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…); L18 (The check `Guide screenshots / Guide images and local UI source review (pull_request)` fai…) New: L22 (Both exclusive paths return before assigning input_in_front, so a skipped focus query now…); L23 (On Linux, a CC value with an unmatched quote raises ValueError in shlex.split during prefl…); L24 (The trailing blank line is flagged by git diff --check. Remove it.); L25 (The fixture ends exclusion and presses G on the same sample, then expects Sprint to turn o…) **Checks:** Still running: `CI / Website build and checks (pull_request)`. <details> <summary>More: does it help, other systems, switches, what was read</summary> **Does it help:** Unproven on 9842f922. The PR reports an earlier Windows build and targeted checks, with no in-game run or frame timings for this head. The cited #1929 measurements cover the earlier Sprint implementation. **Other systems:** The shared Npad path affects keyboard, controllers, Steam Deck, background input, and scripted and bridge input on Windows and Linux. The renderer API bump requires matching host and renderer builds. The diff touches no GPU pass, save format, or audio path. Physical-device and frame-pacing evidence is still missing. **Switches:** Sprint's Controls-page Toggle is saved as toggle_sprint and defaults to Off, preserving held input for existing profiles. This is a player preference, with no new environment switch. The guide documents it. **Reviewed:** 2079d6469f50 against 6e1521a70a32: Reviewed checkout HEAD 9842f922 against ledger head 9482fb1760a5. Read the six changed files through their affected functions and fixtures, the full base-to-HEAD diff, repository rules and PR template, and the relevant menu save, load, Reset, focus, bridge and A-edge code. Skimmed translation tables and screenshot maintenance records. Unchanged portions of large files, screenshot pixels and external Tooling rules were outside the read.; nothing run (a code review). **Checked and fine:** - The basic exclusive A handoff produces a released packet followed by an unmasked A press. A stale physical tap is cleared on a movement-only takeover. - Existing INI files default to held Sprint. Production Reset clears the renderer option and publishes Off to the host. - The renderer API version is bumped to 40, and renderer calls to the new binding are version-gated. - No third-party, generated, shader, game-data, or build-output changes appear in the diff. **Testing:** Read-only review. Both git diff --check comparisons passed. The PR reports earlier Windows llvm-mingw and targeted check results. No builds, tests or game runs were performed here. The supplied CI lists the guide screenshot check as failing and the website check as running. **Overlaps and merge order:** The prerequisite #1948 is merged at the base. No other overlap found in the checkout or supplied review data. </details> --- <sub>Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.</sub>
host/nn_hid.c Outdated
@ -1455,0 +1482,4 @@
/* Sprint's Toggle (the Controls page): a press starts it and the next ends it, ended as the buttons' toggles are
(opening Insert, Plus or Minus, the focus lost) */
const int sprint_toggle = sprint_enabled && sprint_toggle_frame;
const int sprint_front = sprint_toggle && sprint_excluded ? game_in_front() : input_in_front;
Member

L22 (minor): Both exclusive paths return before assigning input_in_front, so a skipped focus query now counts as focus loss. If a movement-only exclusive bridge request runs while the game stays foreground, physical Sprint is cleared and stays off when the request ends. Distinguish an excluded sample from a background sample. Suppress physical input during exclusion and clear the latch on an actual reset.

**L22** (minor): Both exclusive paths return before assigning input_in_front, so a skipped focus query now counts as focus loss. If a movement-only exclusive bridge request runs while the game stays foreground, physical Sprint is cleared and stays off when the request ends. Distinguish an excluded sample from a background sample. Suppress physical input during exclusion and clear the latch on an actual reset.
@ -214,0 +251,4 @@
bridge_up = 1;
bridge_buttons = 1ULL << B_SPRINT; packet(a, 0, 32767, 0);
bridge_buttons = 0; packet(0, 0, 32767, 0);
bridge_active = 0; held_keys['G'] = 1; packet(0, 0, 32767, 0); /* turn the latch off */
Member

L25: The fixture ends exclusion and presses G on the same sample, then expects Sprint to turn off. Exclusion left the latch on and blocked key edges until an unsuppressed sample sees G up. This packet therefore still contains Sprint's A. Sample G released after exclusion ends before pressing it, and assert that the latch is still on at that released sample.

**L25**: The fixture ends exclusion and presses G on the same sample, then expects Sprint to turn off. Exclusion left the latch on and blocked key edges until an unsuppressed sample sees G up. This packet therefore still contains Sprint's A. Sample G released after exclusion ends before pressing it, and assert that the latch is still on at that released sample.
tests/run_all.py Outdated
@ -264,2 +267,2 @@
if a in ("--cc", "--cxx") and not which(args[i + 1]):
missing.append(args[i + 1])
if a in ("--cc", "--cxx"):
try:
Member

L23 (minor): On Linux, a CC value with an unmatched quote raises ValueError in shlex.split during preflight and aborts the suite before checks start. Catch the parse error and report the invalid compiler setting for the affected fixture through the existing preflight result.

**L23** (minor): On Linux, a CC value with an unmatched quote raises ValueError in shlex.split during preflight and aborts the suite before checks start. Catch the parse error and report the invalid compiler setting for the affected fixture through the existing preflight result.
Author
Member
@fix
Member

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is ca90e48b2f. It was already up to date with main.

Changed

  • Exclusive Sprint handoff: host/nn_hid.c:1027 tracks the pending tap's source and preserves an accepted exclusive replay across a second takeover; tests/input_background_test.c:237 checks its press edge.
  • Sprint settings round trip: tests/test_menu_keys.py:265 exercises production save, reload and Controls Reset, including the host-facing value. The fixture was not run.
  • Sprint latch fixture: tests/input_background_test.c:256 samples G released after exclusion, asserts Sprint remains on, then presses G to turn it off.
  • Injected A during exclusive takeover: Preserved queued injected A across takeover while discarding queued physical A, and added the transition to the packet fixture in host/nn_hid.c:1028 and tests/input_background_test.c:250. Tests were not run.
  • Save and Reset fixture cutter: Renamed the local that shadowed block() in tests/test_menu_keys.py:459.
  • Sprint help translations: Added all three strings to the remaining language tables in gpu/vk/menu_i18n.h:116 and extended the check to every table in tests/test_menu_i18n.py:203.

Left for the author

  • Sprint help translations: The remaining language tables and their translation checks still need to be updated; I did not complete the translations.
  • PR summary mask description: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. PR metadata is outside the checkout and the network is closed.
  • PR checklist and results: Clear the full-suite checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description. PR metadata is outside the checkout and the network is closed.
  • Guide screenshot source review: Review the affected guide image, UI sources and capture fixture, then update the review fingerprints or recapture the image as appropriate. The supplied log reports 37 issues but does not contain the complete list; the project checks were not run.
  • PR summary sign-bit explanation: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. The PR description is outside the checkout.
  • PR testing and screenshot checklist: Clear the mise run test checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description. Run the full suite on the updated head before checking it again.
  • Guide screenshot source review: Review the changed UI and guide against settings-graphics.png and the capture fixture, then update the review record and hashes or replace the image as appropriate. Run the guide screenshot checks on the updated head; visual review and capture cannot be completed under the no-run restriction.

Still for the author (no edit does these)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Nothing was built or run: check the change. Its review follows.


Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is ca90e48b2ff2. It was already up to date with `main`. **Changed** - **Exclusive Sprint handoff**: host/nn_hid.c:1027 tracks the pending tap's source and preserves an accepted exclusive replay across a second takeover; tests/input_background_test.c:237 checks its press edge. - **Sprint settings round trip**: tests/test_menu_keys.py:265 exercises production save, reload and Controls Reset, including the host-facing value. The fixture was not run. - **Sprint latch fixture**: tests/input_background_test.c:256 samples G released after exclusion, asserts Sprint remains on, then presses G to turn it off. - **Injected A during exclusive takeover**: Preserved queued injected A across takeover while discarding queued physical A, and added the transition to the packet fixture in host/nn_hid.c:1028 and tests/input_background_test.c:250. Tests were not run. - **Save and Reset fixture cutter**: Renamed the local that shadowed block() in tests/test_menu_keys.py:459. - **Sprint help translations**: Added all three strings to the remaining language tables in gpu/vk/menu_i18n.h:116 and extended the check to every table in tests/test_menu_i18n.py:203. **Left for the author** - **Sprint help translations**: The remaining language tables and their translation checks still need to be updated; I did not complete the translations. - **PR summary mask description**: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. PR metadata is outside the checkout and the network is closed. - **PR checklist and results**: Clear the full-suite checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description. PR metadata is outside the checkout and the network is closed. - **Guide screenshot source review**: Review the affected guide image, UI sources and capture fixture, then update the review fingerprints or recapture the image as appropriate. The supplied log reports 37 issues but does not contain the complete list; the project checks were not run. - **PR summary sign-bit explanation**: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. The PR description is outside the checkout. - **PR testing and screenshot checklist**: Clear the mise run test checkbox, record lint and named base failures, and restore the before-and-after screenshot item in the PR description. Run the full suite on the updated head before checking it again. - **Guide screenshot source review**: Review the changed UI and guide against settings-graphics.png and the capture fixture, then update the review record and hashes or replace the image as appropriate. Run the guide screenshot checks on the updated head; visual review and capture cannot be completed under the no-run restriction. **Still for the author** (no edit does these) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. Nothing was built or run: check the change. Its review follows. --- <sub>Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.</sub>
fix(input): preserve queued Sprint handoffs
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 9s
PR description / Description follows the template (pull_request_target) Successful in 13s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 13s
CI / Workflow lint (actionlint) (pull_request) Successful in 14s
Third-party parts / As their notes say (pull_request) Successful in 15s
CI / Website build and checks (pull_request) Has been cancelled
ca90e48b2f
As the review on #1977 asked.
BeanBot requested changes 2026-11-21 03:08:43 +00:00
Dismissed
BeanBot left a comment

Recommendation: Request changes. The fix preserves a bridge tap that opens Sprint’s A-release gap. An exclusive takeover can still drop a bridge tap that arrives during a gap opened by physical A.

Players can choose press-to-start and press-to-stop Sprint on the Controls page. The setting persists and defaults to held behavior. The feature adds a Sprint latch and tracks the source of queued A presses so exclusive input can discard stale physical input.

@fix will do (4)

  1. L4 host/nn_hid.c:1047: While Sprint supplies A, tap and release physical A, send a one-frame nonexclusive bridge A tap during the 50 ms gap, then start an exclusive movement-only request before the gap expires. No A press reaches the game. Retain injected A as a source when it arrives during a pending gap, and add this sequence to the packet fixture alongside the physical-only discard case.
  2. L14 (minor) gpu/vk/vk_menu.cpp: The code comment is fixed, but the PR summary still says Sprint's index is past the mask's bits. Index 31 is the sign-bit position of a 32-bit int. Update the summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit.
  3. L15 (minor): The PR ticks mise run test but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell.
  4. L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.

For the author (2) (a run, a measurement or a decision)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Since the last review (on 2079d6469f)
Resolved: L9 (The call to block() raises UnboundLocalError on both platforms because main() assign…); L13 (The three new Sprint help strings are translated only in French. A player using Spanish, f…); L25 (The fixture ends exclusion and presses G on the same sample, then expects Sprint to turn o…)
Still open: L4 (While Sprint supplies A, tap and release physical A, send a one-frame nonexclusive bridge…); L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks mise run test but lists only targeted results, and the source shows additio…); L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…); L18 (The check Guide screenshots / Guide images and local UI source review (pull_request) fai…)

Checks: Still running: CI / Website build and checks (pull_request).

More: does it help, other systems, switches, what was read

Does it help: Unproven on this head. The PR reports a Windows build and targeted checks from before the follow-up. There is no in-game result or repeated comparison of average FPS, 1% lows and frame times for the merge base and both head modes. The cited #1929 measurement predates this change.

Other systems: Windows keyboard and XInput, Linux and Deck SDL controllers, background input, host taps and the control bridge converge in the same Npad path. The packet fixture covers injected-only and physical-only takeovers. It misses the two sources arriving during the same gap. No shader, audio or save-format files changed. The host and renderer API version is now 40.

Switches: Sprint’s Controls-page Toggle is saved as toggle_sprint and defaults to Off, preserving held behavior when Sprint is enabled. That default preserves the existing controls. No environment switches changed.

Reviewed: ca90e48b2f against 6e1521a70a: Reviewed checkout ef83bfd896aa against ledger head c96439a2f81e and merge base 6e1521a70a. Read AGENTS.md, CONTRIBUTING.md, the PR template, developer standards, all six follow-up diffs, and the affected code and fixture sections. Traced the input merge, packet writer, menu persistence, and host API. Skimmed the surrounding translation data and guide provenance. The guide PNG and unrelated portions of large files were outside this re-check.; nothing run (a code review).

Checked and fine:

  • Injected-only takeover and physical-only discard are covered by packet assertions on the game-facing button word.
  • The Walk fixture extracts the new input-state declaration with the binding table.
  • The menu uses a separate Sprint field and guards host binding calls by API version.

Testing: Read-only review. git diff --check passed. No builds or tests were run. The supplied CI snapshot reports the guide screenshot source review failing and the website checks still running. Its applicability to ef83bfd8 is unverified.

Overlaps and merge order: #1948 is already in the merge base. No sibling merge dependency was identified in the supplied record.


Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.

**Recommendation: Request changes.** The fix preserves a bridge tap that opens Sprint’s A-release gap. An exclusive takeover can still drop a bridge tap that arrives during a gap opened by physical A. Players can choose press-to-start and press-to-stop Sprint on the Controls page. The setting persists and defaults to held behavior. The feature adds a Sprint latch and tracks the source of queued A presses so exclusive input can discard stale physical input. **`@fix` will do (4)** 1. **L4** `host/nn_hid.c:1047`: While Sprint supplies A, tap and release physical A, send a one-frame nonexclusive bridge A tap during the 50 ms gap, then start an exclusive movement-only request before the gap expires. No A press reaches the game. Retain injected A as a source when it arrives during a pending gap, and add this sequence to the packet fixture alongside the physical-only discard case. 2. **L14** (minor) `gpu/vk/vk_menu.cpp`: The code comment is fixed, but the PR summary still says Sprint's index is past the mask's bits. Index 31 is the sign-bit position of a 32-bit int. Update the summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. 3. **L15** (minor): The PR ticks `mise run test` but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell. 4. **L18**: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s. **For the author (2)** (a run, a measurement or a decision) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. **Since the last review** (on 2079d6469f50) Resolved: L9 (The call to `block()` raises `UnboundLocalError` on both platforms because `main()` assign…); L13 (The three new Sprint help strings are translated only in French. A player using Spanish, f…); L25 (The fixture ends exclusion and presses G on the same sample, then expects Sprint to turn o…) Still open: L4 (While Sprint supplies A, tap and release physical A, send a one-frame nonexclusive bridge…); L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks `mise run test` but lists only targeted results, and the source shows additio…); L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…); L18 (The check `Guide screenshots / Guide images and local UI source review (pull_request)` fai…) **Checks:** Still running: `CI / Website build and checks (pull_request)`. <details> <summary>More: does it help, other systems, switches, what was read</summary> **Does it help:** Unproven on this head. The PR reports a Windows build and targeted checks from before the follow-up. There is no in-game result or repeated comparison of average FPS, 1% lows and frame times for the merge base and both head modes. The cited #1929 measurement predates this change. **Other systems:** Windows keyboard and XInput, Linux and Deck SDL controllers, background input, host taps and the control bridge converge in the same Npad path. The packet fixture covers injected-only and physical-only takeovers. It misses the two sources arriving during the same gap. No shader, audio or save-format files changed. The host and renderer API version is now 40. **Switches:** Sprint’s Controls-page Toggle is saved as toggle_sprint and defaults to Off, preserving held behavior when Sprint is enabled. That default preserves the existing controls. No environment switches changed. **Reviewed:** ca90e48b2ff2 against 6e1521a70a32: Reviewed checkout ef83bfd896aa against ledger head c96439a2f81e and merge base 6e1521a70a32. Read AGENTS.md, CONTRIBUTING.md, the PR template, developer standards, all six follow-up diffs, and the affected code and fixture sections. Traced the input merge, packet writer, menu persistence, and host API. Skimmed the surrounding translation data and guide provenance. The guide PNG and unrelated portions of large files were outside this re-check.; nothing run (a code review). **Checked and fine:** - Injected-only takeover and physical-only discard are covered by packet assertions on the game-facing button word. - The Walk fixture extracts the new input-state declaration with the binding table. - The menu uses a separate Sprint field and guards host binding calls by API version. **Testing:** Read-only review. git diff --check passed. No builds or tests were run. The supplied CI snapshot reports the guide screenshot source review failing and the website checks still running. Its applicability to ef83bfd8 is unverified. **Overlaps and merge order:** #1948 is already in the merge base. No sibling merge dependency was identified in the supplied record. </details> --- <sub>Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.</sub>
Author
Member
@fix
Member

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is cd57d785dd. It was already up to date with main.

Changed

  • Sprint handoff gap, host/nn_hid.c:1047: Retained injected A as a pending source and added the mixed physical/injected sequence beside the physical-only discard case in tests/input_background_test.c:260. Tests were not run.

Left for the author

  • PR summary: Update the PR summary to say the separate Sprint field keeps it out of the Switch-button mask and avoids the sign bit. The PR description is outside the checkout.
  • PR testing checklist: Clear the mise run test checkbox until the full suite runs on the updated head; record lint and the named base failures, and restore the template’s before-and-after screenshot item for the Toggle cell. The PR description is outside the checkout.
  • Guide screenshot source review: Review all 37 reported source and fixture changes against the guide image, update the review record and hashes where justified, then rerun the guide screenshot check. The supplied log contains only the end of the failures, and scripts cannot be run here.
  • PR summary: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. The PR description is not available in the checkout.
  • PR checklist and results: Clear the mise run test checkbox, record lint and the named base failures, and restore the template’s before-and-after screenshot item. The PR description and those results are not available in the checkout.
  • Guide screenshot source review: Review the affected UI, guide and fixture against the recorded screenshot, then update the review hashes or recapture the image as appropriate. The check reports 37 issues; the recorded review predates the checkout’s base, and the required fixture run and visual comparison cannot be performed under this task’s restrictions. No review hashes were changed without that evidence.

Still for the author (no edit does these)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Nothing was built or run: check the change. Its review follows.


Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is cd57d785dd1c. It was already up to date with `main`. **Changed** - **Sprint handoff gap, host/nn_hid.c:1047**: Retained injected A as a pending source and added the mixed physical/injected sequence beside the physical-only discard case in tests/input_background_test.c:260. Tests were not run. **Left for the author** - **PR summary**: Update the PR summary to say the separate Sprint field keeps it out of the Switch-button mask and avoids the sign bit. The PR description is outside the checkout. - **PR testing checklist**: Clear the mise run test checkbox until the full suite runs on the updated head; record lint and the named base failures, and restore the template’s before-and-after screenshot item for the Toggle cell. The PR description is outside the checkout. - **Guide screenshot source review**: Review all 37 reported source and fixture changes against the guide image, update the review record and hashes where justified, then rerun the guide screenshot check. The supplied log contains only the end of the failures, and scripts cannot be run here. - **PR summary**: Update the PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. The PR description is not available in the checkout. - **PR checklist and results**: Clear the `mise run test` checkbox, record lint and the named base failures, and restore the template’s before-and-after screenshot item. The PR description and those results are not available in the checkout. - **Guide screenshot source review**: Review the affected UI, guide and fixture against the recorded screenshot, then update the review hashes or recapture the image as appropriate. The check reports 37 issues; the recorded review predates the checkout’s base, and the required fixture run and visual comparison cannot be performed under this task’s restrictions. No review hashes were changed without that evidence. **Still for the author** (no edit does these) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. Nothing was built or run: check the change. Its review follows. --- <sub>Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.</sub>
fix(input): retain injected taps during Sprint handoffs
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 11s
PR description / Description follows the template (pull_request_target) Successful in 15s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 16s
CI / Workflow lint (actionlint) (pull_request) Successful in 16s
Third-party parts / As their notes say (pull_request) Successful in 19s
CI / Website build and checks (pull_request) Failing after 14s
cd57d785dd
As the review on #1977 asked.
BeanBot left a comment

Recommendation: Request changes. The fix preserves the injected A tap through exclusive takeover. Merge should wait for current-head game runs, held-mode timing comparisons, and the guide screenshot check. (It stands on L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.)

Sprint gains a saved press-to-start, press-to-stop option on the Controls page. The latest change keeps a bridge A tap queued when it arrives during a physical A handoff, at the cost of one source check in the shared input path.

@fix will do (3)

  1. L14 (minor) gpu/vk/vk_menu.cpp: The code comment is fixed, but the PR summary still says Sprint's index is past the mask's bits. Index 31 is the sign-bit position of a 32-bit int. Update the summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit.
  2. L15 (minor): The PR ticks mise run test but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell.
  3. L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.

For the author (2) (a run, a measurement or a decision)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Since the last review (on ca90e48b2f)
Resolved: L4 (While Sprint supplies A, tap and release physical A, send a one-frame nonexclusive bridge…)
Still open: L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks mise run test but lists only targeted results, and the source shows additio…); L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…); L18 (The check Guide screenshots / Guide images and local UI source review (pull_request) fai…)

Checks: Still running: CI / Website build and checks (pull_request).

More: does it help, other systems, switches, what was read

Does it help: Unproven in game. The new fixture covers the 50 ms handoff by inspection, and the PR reports earlier Windows build and targeted test passes. There are no repeated FPS, 1% low, or frame-time results for 6e1521a7 against 8d2c7b53 with the toggle off and on. The #1929 result predates this change.

Other systems: The handoff code is shared by Windows and Linux and handles physical input, scripts, host taps, and the control bridge. The fixture covers exclusive bridge takeover. Windows keyboard and controller behavior and Linux or Steam Deck controller behavior still need game runs. The host and renderer need matching API versions. The guide source review also affects the published Longshot docs.

Switches: Sprint Toggle adds toggle_sprint=, default Off, which preserves hold-to-sprint for existing settings. This is a player input preference. The Controls reset restores Off and the guide documents it. No environment switches changed.

Reviewed: cd57d785dd against 6e1521a70a: Read the repository rules and PR template, the commit log, and the complete ca90e48b-to-8d2c7b53 diff. Read the changed handoff and packet fixture in full with their input routing, callers, packet bounds, and test wiring. Checked the relevant menu, persistence, API, and timing code at HEAD. Skimmed the rest of the base diff, translations, and guide records. The sibling Tooling checkout was unavailable.; nothing run (a code review).

Checked and fine:

  • The exclusive transition preserves a queued injected tap and still drops a physical-only tap.
  • The new fixture checks the emitted A edge and the Sprint press mask. The packet helper also checks both returned states and the output guard bytes.
  • Sprint uses a separate saved field and the host-renderer API version is bumped to 40.

Testing: Read-only review. The latest fix report records no test execution. The PR reports earlier Windows llvm-mingw targeted passes and names test_pads_after_menu as a base failure. The supplied CI reports the guide screenshot check failing only on the head and the website check still running.

Overlaps and merge order: #1948 is already the basis for this follow-up. No other overlap was identified in the supplied material.


Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.

**Recommendation: Request changes.** The fix preserves the injected A tap through exclusive takeover. Merge should wait for current-head game runs, held-mode timing comparisons, and the guide screenshot check. (It stands on L18: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s.) Sprint gains a saved press-to-start, press-to-stop option on the Controls page. The latest change keeps a bridge A tap queued when it arrives during a physical A handoff, at the cost of one source check in the shared input path. **`@fix` will do (3)** 1. **L14** (minor) `gpu/vk/vk_menu.cpp`: The code comment is fixed, but the PR summary still says Sprint's index is past the mask's bits. Index 31 is the sign-bit position of a 32-bit int. Update the summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. 2. **L15** (minor): The PR ticks `mise run test` but lists only targeted results, and the source shows additional fixtures that now cannot compile. Clear that box until the full suite has run on the updated head. Record lint and named base failures, and restore the template’s before-and-after screenshot item for the visible Toggle cell. 3. **L18**: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s. **For the author (2)** (a run, a measurement or a decision) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. **Since the last review** (on ca90e48b2ff2) Resolved: L4 (While Sprint supplies A, tap and release physical A, send a one-frame nonexclusive bridge…) Still open: L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks `mise run test` but lists only targeted results, and the source shows additio…); L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…); L18 (The check `Guide screenshots / Guide images and local UI source review (pull_request)` fai…) **Checks:** Still running: `CI / Website build and checks (pull_request)`. <details> <summary>More: does it help, other systems, switches, what was read</summary> **Does it help:** Unproven in game. The new fixture covers the 50 ms handoff by inspection, and the PR reports earlier Windows build and targeted test passes. There are no repeated FPS, 1% low, or frame-time results for 6e1521a7 against 8d2c7b53 with the toggle off and on. The #1929 result predates this change. **Other systems:** The handoff code is shared by Windows and Linux and handles physical input, scripts, host taps, and the control bridge. The fixture covers exclusive bridge takeover. Windows keyboard and controller behavior and Linux or Steam Deck controller behavior still need game runs. The host and renderer need matching API versions. The guide source review also affects the published Longshot docs. **Switches:** Sprint Toggle adds toggle_sprint=, default Off, which preserves hold-to-sprint for existing settings. This is a player input preference. The Controls reset restores Off and the guide documents it. No environment switches changed. **Reviewed:** cd57d785dd1c against 6e1521a70a32: Read the repository rules and PR template, the commit log, and the complete ca90e48b-to-8d2c7b53 diff. Read the changed handoff and packet fixture in full with their input routing, callers, packet bounds, and test wiring. Checked the relevant menu, persistence, API, and timing code at HEAD. Skimmed the rest of the base diff, translations, and guide records. The sibling Tooling checkout was unavailable.; nothing run (a code review). **Checked and fine:** - The exclusive transition preserves a queued injected tap and still drops a physical-only tap. - The new fixture checks the emitted A edge and the Sprint press mask. The packet helper also checks both returned states and the output guard bytes. - Sprint uses a separate saved field and the host-renderer API version is bumped to 40. **Testing:** Read-only review. The latest fix report records no test execution. The PR reports earlier Windows llvm-mingw targeted passes and names test_pads_after_menu as a base failure. The supplied CI reports the guide screenshot check failing only on the head and the website check still running. **Overlaps and merge order:** #1948 is already the basis for this follow-up. No other overlap was identified in the supplied material. </details> --- <sub>Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.</sub>
Author
Member
@fix
Member

I didn't change this pull request: what is left needs the author. It was already up to date with main.

Left for the author

  • PR summary: Edit the Forgejo PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. The PR description is not in the checkout.
  • PR checklist and results: In the Forgejo PR description, clear the full-suite checkbox, record the actual lint result and named base failures, and restore the before-and-after screenshot item. Run the full suite on the updated head before checking its box.
  • Guide screenshot source review: Review all 37 reported source and fixture changes against the affected guide images, recapture where needed, and update the review evidence and hashes. The required rendering and checks cannot be run under this task’s restrictions; refreshing hashes without that review would misrepresent the evidence.

Still for the author (no edit does these)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Nothing was pushed.


Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.

I didn't change this pull request: what is left needs the author. It was already up to date with `main`. **Left for the author** - **PR summary**: Edit the Forgejo PR summary to say the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. The PR description is not in the checkout. - **PR checklist and results**: In the Forgejo PR description, clear the full-suite checkbox, record the actual lint result and named base failures, and restore the before-and-after screenshot item. Run the full suite on the updated head before checking its box. - **Guide screenshot source review**: Review all 37 reported source and fixture changes against the affected guide images, recapture where needed, and update the review evidence and hashes. The required rendering and checks cannot be run under this task’s restrictions; refreshing hashes without that review would misrepresent the evidence. **Still for the author** (no edit does these) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. Nothing was pushed. --- <sub>Automated fix by BeanBot (gpt-6.1-terra), asked for by llun. Nothing was run; a maintainer decides.</sub>
fix(controls): clarify Sprint and stick bindings
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 9s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 11s
CI / Workflow lint (actionlint) (pull_request) Successful in 13s
Third-party parts / As their notes say (pull_request) Successful in 13s
CI / Website build and checks (pull_request) Failing after 16s
PR description / Description follows the template (pull_request_target) Successful in 7s
cbfa7a1811
Show the gamepad stick clicks as L3 and R3, place Sprint after Walk, and explain the L3 assignment and Sprint modes. Keep action labels and bindings intact.

Tested the Windows renderer in game and passed all 12 menu translation checks.

Assisted-by: OpenAI Codex

Co-Authored-By: OpenAI Codex <[email protected]>
Owner
@fix
Member

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is 9b662678d3. I rebased it onto main (ed602b9468). I edited the pull request's description too (5 edits).

Changed

  • Sprint help, gpu/vk/vk_menu.cpp:2570: Clarified the L3 primary-binding condition and updated all corresponding translations in gpu/vk/menu_i18n.h:197. No tests were run.
  • Sprint mask explanation: Prepared a summary edit explaining that the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit.

Changed in part (the rest is for the author)

  • Test and screenshot checklist: Prepared edits to clear the suite checkbox, restore the screenshot checkbox, and record that HEAD suite and lint results are pending. The author must run the full suite and lint on the updated head, record their results and named base failures, and provide the before-and-after screenshots.

Left for the author

  • Guide screenshot source review: Compare the affected screenshot and guide with the current UI, then update the reviewed hashes or recapture the image as needed. The log reports 37 issues but includes only the final three; no visual comparison was run.
  • Website build failure: Provide the error message preceding the stack trace or rerun the website check and inspect its full output. The supplied log identifies the throw site but omits the error and affected setting.

Still for the author (no edit does these)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.
  • L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.
  • L27: The check CI / Website build and checks (pull_request) fails on this head and passes on the base branch: Failing after 16s.

Nothing was built or run: check the change. It will be reviewed again.


Automated fix by BeanBot (gpt-6.1-terra), asked for by HumanBean. Nothing was run; a maintainer decides.

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is 9b662678d321. I rebased it onto `main` (ed602b946814). I edited the pull request's description too (5 edits). **Changed** - **Sprint help, gpu/vk/vk_menu.cpp:2570**: Clarified the L3 primary-binding condition and updated all corresponding translations in gpu/vk/menu_i18n.h:197. No tests were run. - **Sprint mask explanation**: Prepared a summary edit explaining that the separate field keeps Sprint out of the Switch-button mask and avoids its sign bit. **Changed in part (the rest is for the author)** - **Test and screenshot checklist**: Prepared edits to clear the suite checkbox, restore the screenshot checkbox, and record that HEAD suite and lint results are pending. The author must run the full suite and lint on the updated head, record their results and named base failures, and provide the before-and-after screenshots. **Left for the author** - **Guide screenshot source review**: Compare the affected screenshot and guide with the current UI, then update the reviewed hashes or recapture the image as needed. The log reports 37 issues but includes only the final three; no visual comparison was run. - **Website build failure**: Provide the error message preceding the stack trace or rerun the website check and inspect its full output. The supplied log identifies the throw site but omits the error and affected setting. **Still for the author** (no edit does these) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. - **L18**: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s. - **L27**: The check `CI / Website build and checks (pull_request)` fails on this head and passes on the base branch: Failing after 16s. Nothing was built or run: check the change. It will be reviewed again. --- <sub>Automated fix by BeanBot (gpt-6.1-terra), asked for by HumanBean. Nothing was run; a maintainer decides.</sub>
BeanBot force-pushed feat/sprint-toggle from cbfa7a1811
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 9s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 11s
CI / Workflow lint (actionlint) (pull_request) Successful in 13s
Third-party parts / As their notes say (pull_request) Successful in 13s
CI / Website build and checks (pull_request) Failing after 16s
PR description / Description follows the template (pull_request_target) Successful in 7s
to 9b662678d3
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 18s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 22s
CI / Website build and checks (pull_request) Failing after 23s
CI / Workflow lint (actionlint) (pull_request) Successful in 23s
Third-party parts / As their notes say (pull_request) Successful in 25s
PR description / Description follows the template (pull_request_target) Successful in 23s
2026-11-21 06:01:10 +00:00
Compare
BeanBot left a comment

Recommendation: Request changes. The Sprint Toggle preserves held mode as the default and adds a useful input choice. Merge should wait for physical-input coverage, both-mode frame timings, matching Controls captures, and passing head checks. (It stands on L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.)

Players can press once to start Sprint and again to stop it, with the choice saved across launches. The latest commits put Sprint next to Walk and label physical stick clicks L3 and R3. The feature adds one saved setting and host state for the latch and its handoffs.

@fix will do (4)

  1. L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.
  2. L26 (minor) gpu/vk/vk_menu.cpp:2559: The new Left stick press hint checks Sprint's stored bindings alone. (more on the line)
  3. L27 (minor): The summary and Risks omit visible changes with Sprint Toggle Off. For example, the Xbox binding cells now show L3 and R3 and Sprint appears after Walk. Mention the new labels, order, and binding help, and keep one copy of the same-spot screenshots checklist line.
  4. L28: The check CI / Website build and checks (pull_request) fails on this head and passes on the base branch: Failing after 23s.

For the author (2) (a run, a measurement or a decision)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Since the last review (on cd57d785dd)
Resolved: L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks mise run test but lists only targeted results, and the source shows additio…)
Still open: L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…); L18 (The check Guide screenshots / Guide images and local UI source review (pull_request) fai…)
New: L26 (The new Left stick press hint checks Sprint's stored bindings alone. If the player also as…); L27 (The summary and Risks omit visible changes with Sprint Toggle Off. For example, the Xbox b…); L28 (The check CI / Website build and checks (pull_request) fails on this head and passes on…)

More: does it help, other systems, switches, what was read

Does it help: Unproven on the current head. Earlier Windows fixture passes and an in-game renderer smoke test are reported. There are no repeated interleaved runs of the base and both head modes with average FPS, 1% lows, and frame times. The #1929 comparison predates this Toggle.

Other systems: The shared host input path affects keyboard and controller on Windows and Linux, including Steam Deck, and shares A-button handling with physical input, scripts, the control bridge, and the ocarina. Fixtures cover focus and injected A handoffs. The latest renderer changes affect Controls labels and layout across prompt styles and languages. The diff leaves save data, audio, and GPU submission paths untouched.

Switches: Sprint Toggle defaults Off, keeps Sprint held while enabled, and resets to Off with Controls defaults. That preserves the existing input behavior by default. The separate Sprint enable setting still defaults Off. No runtime environment switches were added or changed.

Reviewed: 9b662678d3 against ed602b9468: Read AGENTS.md, CONTRIBUTING.md, the Forgejo PR template, and tests/test_menu_i18n.py in full. Read the changed regions of vk_menu.cpp and nn_hid.c, the Sprint module, and related binding, save, load, reset, and host API code. Reviewed fixture changes and skimmed translation entries and reported runs. No docs/review-checklist.md was present. Unchanged portions of large files and individual translations were not exhaustively reviewed.; nothing run (a code review).

Checked and fine:

  • Moving Sprint after Walk changes display order without renumbering its binding ID.
  • The new Sprint hint describes the two empty Sprint pad slots and primary left-stick binding required by the automatic L3 move.
  • The translation check requires the updated help and action labels in every table.
  • Host API version 40 gates the new binding operation. The diff adds no game files, build output, shader changes, or third-party changes.

Testing: Read-only review. The Windows build and targeted fixture passes come from the PR description. Commit c786908c reports a Windows in-game renderer smoke test and 12 passing menu translation checks before the final wording change. The full suite and lint are pending. Supplied CI reports head-only failures for Guide screenshots and Website build and checks.

Overlaps and merge order: #1948 is the merged Sprint predecessor. No pending overlap or merge-order dependency is identified in the supplied material.


Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.

**Recommendation: Request changes.** The Sprint Toggle preserves held mode as the default and adds a useful input choice. Merge should wait for physical-input coverage, both-mode frame timings, matching Controls captures, and passing head checks. (It stands on L18: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s.) Players can press once to start Sprint and again to stop it, with the choice saved across launches. The latest commits put Sprint next to Walk and label physical stick clicks L3 and R3. The feature adds one saved setting and host state for the latch and its handoffs. **`@fix` will do (4)** 1. **L18**: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s. 2. **L26** (minor) `gpu/vk/vk_menu.cpp:2559`: The new Left stick press hint checks Sprint's stored bindings alone. (more on the line) 3. **L27** (minor): The summary and Risks omit visible changes with Sprint Toggle Off. For example, the Xbox binding cells now show L3 and R3 and Sprint appears after Walk. Mention the new labels, order, and binding help, and keep one copy of the same-spot screenshots checklist line. 4. **L28**: The check `CI / Website build and checks (pull_request)` fails on this head and passes on the base branch: Failing after 23s. **For the author (2)** (a run, a measurement or a decision) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. **Since the last review** (on cd57d785dd1c) Resolved: L14 (The code comment is fixed, but the PR summary still says Sprint's index is past the mask's…); L15 (The PR ticks `mise run test` but lists only targeted results, and the source shows additio…) Still open: L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…); L18 (The check `Guide screenshots / Guide images and local UI source review (pull_request)` fai…) New: L26 (The new Left stick press hint checks Sprint's stored bindings alone. If the player also as…); L27 (The summary and Risks omit visible changes with Sprint Toggle Off. For example, the Xbox b…); L28 (The check `CI / Website build and checks (pull_request)` fails on this head and passes on…) <details> <summary>More: does it help, other systems, switches, what was read</summary> **Does it help:** Unproven on the current head. Earlier Windows fixture passes and an in-game renderer smoke test are reported. There are no repeated interleaved runs of the base and both head modes with average FPS, 1% lows, and frame times. The #1929 comparison predates this Toggle. **Other systems:** The shared host input path affects keyboard and controller on Windows and Linux, including Steam Deck, and shares A-button handling with physical input, scripts, the control bridge, and the ocarina. Fixtures cover focus and injected A handoffs. The latest renderer changes affect Controls labels and layout across prompt styles and languages. The diff leaves save data, audio, and GPU submission paths untouched. **Switches:** Sprint Toggle defaults Off, keeps Sprint held while enabled, and resets to Off with Controls defaults. That preserves the existing input behavior by default. The separate Sprint enable setting still defaults Off. No runtime environment switches were added or changed. **Reviewed:** 9b662678d321 against ed602b946814: Read AGENTS.md, CONTRIBUTING.md, the Forgejo PR template, and tests/test_menu_i18n.py in full. Read the changed regions of vk_menu.cpp and nn_hid.c, the Sprint module, and related binding, save, load, reset, and host API code. Reviewed fixture changes and skimmed translation entries and reported runs. No docs/review-checklist.md was present. Unchanged portions of large files and individual translations were not exhaustively reviewed.; nothing run (a code review). **Checked and fine:** - Moving Sprint after Walk changes display order without renumbering its binding ID. - The new Sprint hint describes the two empty Sprint pad slots and primary left-stick binding required by the automatic L3 move. - The translation check requires the updated help and action labels in every table. - Host API version 40 gates the new binding operation. The diff adds no game files, build output, shader changes, or third-party changes. **Testing:** Read-only review. The Windows build and targeted fixture passes come from the PR description. Commit c786908c reports a Windows in-game renderer smoke test and 12 passing menu translation checks before the final wording change. The full suite and lint are pending. Supplied CI reports head-only failures for Guide screenshots and Website build and checks. **Overlaps and merge order:** #1948 is the merged Sprint predecessor. No pending overlap or merge-order dependency is identified in the supplied material. </details> --- <sub>Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.</sub>
@ -2541,19 +2556,28 @@ void draw_controls(ui::Ctx& c) {
const ui::Prompt wait[] = {{ui::kBack, "Cancel"}};
const float bar_top = ui::prompt_bar_top(c); // (the strip itself drawn last, over all: below)
const Row* page_row = sel_i < 0 ? &rows[tab_row_to_global(TAB_PAGE, csel)] : nullptr;
const bool l3_sprints = opts.sprint && sel_i >= 0 && sel_i < kControls && controls[sel_i].button == HOS_SW_LS &&
Member

L26 (minor): The new Left stick press hint checks Sprint's stored bindings alone. If the player also assigns L3 to Toggle HUD, the host keeps both assignments but masks L3 out of Sprint. The hint still tells the player L3 is assigned to Sprint without explaining why it fails. Check both Toggle HUD pad slots before showing the Sprint-specific hint.

**L26** (minor): The new Left stick press hint checks Sprint's stored bindings alone. If the player also assigns L3 to Toggle HUD, the host keeps both assignments but masks L3 out of Sprint. The hint still tells the player L3 is assigned to Sprint without explaining why it fails. Check both Toggle HUD pad slots before showing the Sprint-specific hint.
Owner
@fix
Member

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is 07d779e387. I rebased it onto main (40bfceb055).

Conflicts resolved

  • docs/user/graphics-and-controls.md: Kept the Sprint toggle explanation and the base branch’s instrument wording. I edited the pull request's description too (5 edits).

Changed

  • Sprint L3 hint: gpu/vk/vk_menu.cpp:2559 now checks both Toggle HUD pad slots before showing the Sprint-specific hint.
  • Visible Controls changes and duplicate checklist line: Prepared description edits covering the labels, order and binding help, and removing the duplicate screenshot line.
  • Graphics guide repair in summary: Prepared a description edit covering Reset HUD placement, Custom HUD visibility and website test coverage.

Already done

  • Website build failure: The committed repair in website/src/repo-docs/graphics-settings.ts:25 and :256 recognizes the Custom HUD condition and Reset HUD placement action. The website check has not been rerun here.

Left for the author

  • Guide screenshot source review: After the branch is updated, review the affected screenshot and guide against the merged UI and fixture sources, update the review record and hashes as appropriate, and rerun the check. The base has replaced the screenshot and its review record, so the 37 reported mismatches need to be assessed on the updated tree.

Still for the author (no edit does these)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.
  • L18: The check Guide screenshots / Guide images and local UI source review (pull_request) fails on this head and passes on the base branch: Failing after 11s.

Nothing was built or run: check the change. It will be reviewed again.


Automated fix by BeanBot (gpt-6.1-terra), asked for by HumanBean. Nothing was run; a maintainer decides.

I changed this pull request where an edit could (the rest needs runs, measurements or decisions): the new head is 07d779e3874d. I rebased it onto `main` (40bfceb0552e). **Conflicts resolved** - docs/user/graphics-and-controls.md: Kept the Sprint toggle explanation and the base branch’s instrument wording. I edited the pull request's description too (5 edits). **Changed** - **Sprint L3 hint**: gpu/vk/vk_menu.cpp:2559 now checks both Toggle HUD pad slots before showing the Sprint-specific hint. - **Visible Controls changes and duplicate checklist line**: Prepared description edits covering the labels, order and binding help, and removing the duplicate screenshot line. - **Graphics guide repair in summary**: Prepared a description edit covering Reset HUD placement, Custom HUD visibility and website test coverage. **Already done** - **Website build failure**: The committed repair in website/src/repo-docs/graphics-settings.ts:25 and :256 recognizes the Custom HUD condition and Reset HUD placement action. The website check has not been rerun here. **Left for the author** - **Guide screenshot source review**: After the branch is updated, review the affected screenshot and guide against the merged UI and fixture sources, update the review record and hashes as appropriate, and rerun the check. The base has replaced the screenshot and its review record, so the 37 reported mismatches need to be assessed on the updated tree. **Still for the author** (no edit does these) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. - **L18**: The check `Guide screenshots / Guide images and local UI source review (pull_request)` fails on this head and passes on the base branch: Failing after 11s. Nothing was built or run: check the change. It will be reviewed again. --- <sub>Automated fix by BeanBot (gpt-6.1-terra), asked for by HumanBean. Nothing was run; a maintainer decides.</sub>
BeanBot force-pushed feat/sprint-toggle from 9b662678d3
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 18s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 22s
CI / Website build and checks (pull_request) Failing after 23s
CI / Workflow lint (actionlint) (pull_request) Successful in 23s
Third-party parts / As their notes say (pull_request) Successful in 25s
PR description / Description follows the template (pull_request_target) Successful in 23s
to 07d779e387
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 27s
PR description / Description follows the template (pull_request_target) Successful in 47s
Guide screenshots / Guide images and local UI source review (pull_request) Failing after 42s
CI / Workflow lint (actionlint) (pull_request) Successful in 45s
Third-party parts / As their notes say (pull_request) Successful in 45s
CI / Website build and checks (pull_request) Has been cancelled
2026-11-21 06:18:42 +00:00
Compare
Read the new host snapshots in the fixture while proving Walk polls do not enter a Sprint toggle or injection path.

Assisted-by: OpenAI Codex (GPT-6.2)
docs: verify Graphics guide against the sprint menu
Some checks failed
CI / Python source checks (ubuntu-24.04, part 1/2) (pull_request) Has been skipped
CI / Python source checks (ubuntu-24.04, part 2/2) (pull_request) Has been skipped
CI / SPIR-V fixtures (global narrow load/store, direct surface atomics, F2I immediate) (pull_request) Has been skipped
Third-party parts / The same as upstream (pull_request) Has been skipped
Website / Website on Cloudflare (pull_request) Has been skipped
CI / Conventional Commit title (pull_request) Successful in 11s
CI / Workflow lint (actionlint) (pull_request) Successful in 21s
Guide screenshots / Guide images and local UI source review (pull_request) Successful in 24s
Third-party parts / As their notes say (pull_request) Successful in 26s
PR description / Description follows the template (pull_request_target) Failing after 14s
CI / Website build and checks (pull_request) Successful in 2m23s
cfa14d6eba
Retain the Graphics image after matching its content against the current main and Sprint captures. The later left-stick help condition only draws on the Controls binding page.

Assisted-by: OpenAI Codex (GPT-6.2)
HumanBean changed title from WIP: feat(input): a Toggle for Sprint on the Controls page, Off (held) by default to feat(input): a Toggle for Sprint on the Controls page, Off (held) by default 2026-11-21 06:20:31 +00:00
HumanBean deleted branch feat/sprint-toggle 2026-11-21 06:20:48 +00:00
Member

The branch is current with main 40bfceb0, including the website and Graphics guide fix from #2013. The latest update fixes the Walk fixture on Windows llvm-mingw and records the Graphics source comparison. The Graphics view matched the existing guide image outside the version label, so the image stays as captured.

The Controls images below render the same selected Sprint binding at 1280×720 with synthetic Shift and left-stick bindings. They show the move under Walk, L3 and the new Toggle cell with its default Off state. These are game-free UI captures. The PR image uses the 2a74b3f2 menu source. The only subsequent rendering change is a help condition shown when the separate left-stick row is selected.

Before · main 40bfceb0

Sprint Controls on main

After · Sprint Toggle

Sprint Controls with the Sprint Toggle cell

The branch is current with main `40bfceb0`, including the website and Graphics guide fix from #2013. The latest update fixes the Walk fixture on Windows llvm-mingw and records the Graphics source comparison. The Graphics view matched the existing guide image outside the version label, so the image stays as captured. The Controls images below render the same selected Sprint binding at 1280×720 with synthetic Shift and left-stick bindings. They show the move under Walk, L3 and the new Toggle cell with its default Off state. These are game-free UI captures. The PR image uses the `2a74b3f2` menu source. The only subsequent rendering change is a help condition shown when the separate left-stick row is selected. **Before** · main `40bfceb0` ![Sprint Controls on main](https://git.magicbean.dev/attachments/b6e198bd-5c5b-49a1-82fa-ffa0e45ca497) **After** · Sprint Toggle ![Sprint Controls with the Sprint Toggle cell](https://git.magicbean.dev/attachments/240c37c8-edb7-401d-b0ce-724cb2f27238)
BeanBot left a comment

Recommendation: Wait. The L3 hint now respects both Toggle HUD bindings, and the PR description covers the visible Controls changes. Merge readiness still depends on physical input runs and the matched timing comparison in the ledger.

Sprint gains an opt-in press-to-toggle mode while retaining held input by default. The change reuses the Controls page’s Toggle column and adds a saved setting, host latch state, and A-button handoff logic.

@fix will do (2)

  1. L29 (minor) website/tests/graphics-settings.test.ts:35: The new test repeats HUD visibility and reset-hook coverage already present on the rebased base in website/tests/hud-visibility.test.ts. (more on the line)
  2. L30 (minor): The Summary credits this PR with repairing Graphics guide support for Reset HUD placement. A reader comparing against 40bfceb055 will find the reset parser and Custom HUD visibility text already on the base. Remove that claim so the Summary matches the rebased diff.

For the author (2) (a run, a measurement or a decision)

  • L16 (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison.
  • L17 (minor) host/nn_hid.c:1245: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times.

Since the last review (on 9b662678d3)
Resolved: L18 (The check Guide screenshots / Guide images and local UI source review (pull_request) fai…); L26 (The new Left stick press hint checks Sprint's stored bindings alone. If the player also as…); L27 (The summary and Risks omit visible changes with Sprint Toggle Off. For example, the Xbox b…); L28 (The check CI / Website build and checks (pull_request) fails on this head and passes on…)
Still open: L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…)
New: L29 (The new test repeats HUD visibility and reset-hook coverage already present on the rebased…); L30 (The Summary credits this PR with repairing Graphics guide support for Reset HUD placement.…)

Checks: Still running: PR description / Description follows the template (pull_request_target).

More: does it help, other systems, switches, what was read

Does it help: Unproven in game. The reported Windows build and targeted tests support the toggle logic, but the Windows renderer smoke test does not establish physical input behavior. The #1929 numbers cover the earlier Sprint implementation. This PR has no repeated base/head FPS, 1% low, or frame-time results for its two modes.

Other systems: The host changes affect keyboard, controller, focus and background input, scripted input, and the control bridge. The fixtures cover focus changes and injected A taps during handoffs. Physical controller behavior on Linux and Steam Deck and applicable Tier 1 coverage remain unverified. The website’s Graphics guide parser and HUD visibility documentation came in through the rebased base.

Switches: The player setting toggle_sprint defaults to 0, which preserves held Sprint when Sprint is enabled. That is the right compatibility default. The guide documents it and the menu fixture covers persistence and reset. No environment switch changes.

Reviewed: 07d779e387 against 40bfceb055: Read AGENTS.md, CONTRIBUTING.md, the Forgejo PR template, and the hardware tiers in docs/dev/standards.md. Read the changed-since hunks for vk_menu.cpp, the player guide, and the website test, plus the host input and website parser code those hunks rely on. Skimmed the remaining PR diff, translations, commit history, and reported runs. The unchanged files from earlier rounds were not reread in full. No docs/review-checklist.md exists in the checkout.; nothing run (a code review).

Checked and fine:

  • The L3 hint checks both Toggle HUD slots, matching the host’s input precedence.
  • The rebased guide already documents the Custom HUD condition, and the parser already recognizes the registered reset action.
  • The Sprint guide still distinguishes enabling Sprint from changing its input between held and toggle.

Testing: Read-only review. The PR reports a Windows llvm-mingw build and earlier targeted fixture passes. Updated-head full-suite and lint results, physical input runs, and timing measurements are still outstanding. The supplied CI snapshot lists the description check as running and no failures.

Overlaps and merge order: #1948 is the Sprint prerequisite and is already in the base. The rebased base also contains the Graphics guide repair. No sibling merge dependency was identified.


Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.

**Recommendation: Wait.** The L3 hint now respects both Toggle HUD bindings, and the PR description covers the visible Controls changes. Merge readiness still depends on physical input runs and the matched timing comparison in the ledger. Sprint gains an opt-in press-to-toggle mode while retaining held input by default. The change reuses the Controls page’s Toggle column and adds a saved setting, host latch state, and A-button handoff logic. **`@fix` will do (2)** 1. **L29** (minor) `website/tests/graphics-settings.test.ts:35`: The new test repeats HUD visibility and reset-hook coverage already present on the rebased base in website/tests/hud-visibility.test.ts. (more on the line) 2. **L30** (minor): The Summary credits this PR with repairing Graphics guide support for Reset HUD placement. A reader comparing against 40bfceb0552e will find the reset parser and Custom HUD visibility text already on the base. Remove that claim so the Summary matches the rebased diff. **For the author (2)** (a run, a measurement or a decision) - **L16** (minor): The reported build and game-free tests leave the physical input path unverified. Record a base/head game run with keyboard and controller on Windows and controller on Linux or Steam Deck, including release while steering, A handoff, resets, rebinding, and persistence. Include applicable Tier 1 results, matching menu captures, and the guide source comparison. - **L17** (minor) `host/nn_hid.c:1245`: The held path now skips the latch helper and still loads the toggle snapshot. Run the base and both head modes on the same hardware, scene, settings and build type, with repeated interleaved runs. Report average FPS, 1% lows and frame times. **Since the last review** (on 9b662678d321) Resolved: L18 (The check `Guide screenshots / Guide images and local UI source review (pull_request)` fai…); L26 (The new Left stick press hint checks Sprint's stored bindings alone. If the player also as…); L27 (The summary and Risks omit visible changes with Sprint Toggle Off. For example, the Xbox b…); L28 (The check `CI / Website build and checks (pull_request)` fails on this head and passes on…) Still open: L16 (The reported build and game-free tests leave the physical input path unverified. Record a…); L17 (The held path now skips the latch helper and still loads the toggle snapshot. Run the base…) New: L29 (The new test repeats HUD visibility and reset-hook coverage already present on the rebased…); L30 (The Summary credits this PR with repairing Graphics guide support for Reset HUD placement.…) **Checks:** Still running: `PR description / Description follows the template (pull_request_target)`. <details> <summary>More: does it help, other systems, switches, what was read</summary> **Does it help:** Unproven in game. The reported Windows build and targeted tests support the toggle logic, but the Windows renderer smoke test does not establish physical input behavior. The #1929 numbers cover the earlier Sprint implementation. This PR has no repeated base/head FPS, 1% low, or frame-time results for its two modes. **Other systems:** The host changes affect keyboard, controller, focus and background input, scripted input, and the control bridge. The fixtures cover focus changes and injected A taps during handoffs. Physical controller behavior on Linux and Steam Deck and applicable Tier 1 coverage remain unverified. The website’s Graphics guide parser and HUD visibility documentation came in through the rebased base. **Switches:** The player setting `toggle_sprint` defaults to 0, which preserves held Sprint when Sprint is enabled. That is the right compatibility default. The guide documents it and the menu fixture covers persistence and reset. No environment switch changes. **Reviewed:** 07d779e3874d against 40bfceb0552e: Read AGENTS.md, CONTRIBUTING.md, the Forgejo PR template, and the hardware tiers in docs/dev/standards.md. Read the changed-since hunks for vk_menu.cpp, the player guide, and the website test, plus the host input and website parser code those hunks rely on. Skimmed the remaining PR diff, translations, commit history, and reported runs. The unchanged files from earlier rounds were not reread in full. No docs/review-checklist.md exists in the checkout.; nothing run (a code review). **Checked and fine:** - The L3 hint checks both Toggle HUD slots, matching the host’s input precedence. - The rebased guide already documents the Custom HUD condition, and the parser already recognizes the registered reset action. - The Sprint guide still distinguishes enabling Sprint from changing its input between held and toggle. **Testing:** Read-only review. The PR reports a Windows llvm-mingw build and earlier targeted fixture passes. Updated-head full-suite and lint results, physical input runs, and timing measurements are still outstanding. The supplied CI snapshot lists the description check as running and no failures. **Overlaps and merge order:** #1948 is the Sprint prerequisite and is already in the base. The rebased base also contains the Graphics guide repair. No sibling merge dependency was identified. </details> --- <sub>Automated code review by BeanBot (gpt-6.2-sol), asked for by BeanBot, after its fix. Nothing was run; a maintainer decides.</sub>
@ -32,6 +32,17 @@ test('CRT-line visibility has a badge and still requires its documented conditio
assert.throws(() => readGraphicsSettings(undocumented, menu), /no condition for CRT lines/);
});
test('Custom HUD rows require their visibility condition and the registered reset action', () => {
Member

L29 (minor): The new test repeats HUD visibility and reset-hook coverage already present on the rebased base in website/tests/hud-visibility.test.ts. For example, renaming the registered reset hook now triggers the same coverage assertion in both files. Drop the duplicate test and leave the existing HUD test as the single place for these cases.

**L29** (minor): The new test repeats HUD visibility and reset-hook coverage already present on the rebased base in website/tests/hud-visibility.test.ts. For example, renaming the registered reset hook now triggers the same coverage assertion in both files. Drop the duplicate test and leave the existing HUD test as the single place for these cases.
Sign in to join this conversation.
No reviewers
No milestone
No project
No assignees
4 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Longshot!1977
No description provided.