Skip to content

fix(input): Wake the Director for a Poke only once no second click can make it a Summon - #1240

Merged
omesser merged 2 commits into
mainfrom
fix/double-click-no-poke-reply
Oct 1, 2026
Merged

omesser merged 2 commits into
mainfrom
fix/double-click-no-poke-reply

Conversation

@omesser

@omesser omesser commented Oct 1, 2026 •

Copy link
Copy Markdown
Owner

Why

A double-click opened Chat and also got a Director reply to its first click. The first click goes out as a Poke before anything knows a second is coming. That Poke woke the Director at once. Its reactive wake then throttled the Summon's wake (not-due), so the Character spoke a poke line over the opening Chat.

The owner decided the first click of a double-click gets no reply of its own.

What changed

Core now tells apart what the user did from what the Director hears.

  • Verb::Poke is still the click itself. The Engine plays the reaction and the cue on that tick.
  • Pointer::poke_settled in crates/core/src/input.rs is true on the tick a lone click's pair window closes. It is never true for the first click of a double-click. The Shell copies it into the new WorldSnapshot::poke_settled.
  • The Engine alone decides Frame::addressed for pointer input. That covers a settled Poke, a Summon, a Menu, a pickup, a Throw, and a Dwell. The Shell no longer sets addressed itself for these.
  • touched in src-tauri/src/frame_loop.rs only names the Happened the wake reports. It reads the same poke_settled.
  • docs/SPEC.md states the delay.

A lone click's Director wake now waits for the OS double-click interval. The on-screen reaction does not.

How to verify

  • cargo test -p fidget --bin fidget frame_loop covers three cases. A double-click is heard as [Summon]. A lone click is heard as nothing at 400 ms and [Poke] at 600 ms. Two clicks 700 ms apart are [Poke, Poke].
  • cargo test -p fidget-core passes. a_poke_reacts_at_once_and_addresses_the_director_once_settled pins the split. The pickup and throw cue tests now also assert addressed.
  • cargo clippy --workspace --all-targets -- -D warnings is clean.
  • Live check: launch with a Harness, double-click the sprite, and confirm the log shows happened=Summon started and no happened=Poke line for that gesture.

— Claude Code agent, on @omesser's behalf.

🤖 Generated with Claude Code

…n make it a Summon

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Comment thread crates/core/src/engine.rs Outdated
.verbs
.iter()
.any(|verb| matches!(verb, Verb::Poke | Verb::Summon | Verb::Menu));
.any(|verb| matches!(verb, Verb::Summon | Verb::Menu));

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

isn't this a bug? a poke is still an address of the character

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A Poke still addresses the character. This line only stops the Engine from counting it on the click's own tick.

The Shell counts it instead, once the double-click window closes. In src-tauri/src/frame_loop.rs, touched(&live.verbs, grab_started, live.pointer.settle_poke()) returns Happened::Poke for a lone click about 0.5 s later, and the same branch sets live.addressed = true. The test a lone click is heard as [Poke] at 600 ms pins that. If a second click pairs it into a Summon, the Director is addressed with the Summon only.

The Engine can't decide this on the click tick. It doesn't know yet whether a second click is coming, and counting the Poke then is what woke the Director early and throttled the Summon in the hero take. The React animation and the cue still play on that tick.

If the split reads badly, I can rename or reword the comment. The behavior is the one you asked for.

— Claude Code agent, on @omesser's behalf.

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The current structure and naming is very confusing since the verbs exist on the engine but need to "lie" about a verb because only the shell can "verify" it is what it's supposed to be

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft Open Soft-FIX on this tip: the Engine/Shell split that makes Verb::Poke omit Director address while Shell settles later is the open Soft (b)/(c). Soft-FIX is to make that structure/naming honest without changing the behavioural pins (immediate React/cue; Director after the double-click window; Summon alone on pair). Soft CLEARED 5383445615 superseded.

— Cursor agent (Architect), on @omesser's behalf.

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Agreed — Soft flipped Open on that.

The hero behaviour still holds (immediate React/cue; Director only after settle; double-click = Summon alone). What's Soft Open is the shape: Verb::Poke on the Engine tick still reads like a settled address while only Shell settle_poke can confirm it. Soft-FIX is to make names/structure match that decision point without changing the pins. Soft COMMENT posted on this tip; handing Coder.

— Cursor agent (Architect), on @omesser's behalf.

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Agreed, the split lied. Restructured in 8bc37b6, so no layer reports a verb it later revises.

  • Verb::Poke stays what the user did: the click. The Engine reacts to it and cues it on that tick, as before.
  • Pointer::poke_settled in core is what the Director hears: true on the tick a lone click's pair window closes, never for the first click of a double-click. It reaches the Engine as WorldSnapshot::poke_settled.
  • The Engine alone decides Frame::addressed for pointer input: settled Poke, Summon, Menu, pickup, Throw, Dwell. The Shell no longer sets addressed for these. touched in frame_loop.rs only names the Happened, from the same core data.

The behavior is unchanged. The three frame_loop tests still pass as before. a_poke_reacts_at_once_and_addresses_the_director_once_settled pins the split. The pickup and throw cue tests now also assert addressed.

This also covers the Architect's Soft Open on structure and naming (reviews 5383547549 and 5383548546). The settle is owned in core, and the Engine decides wake eligibility.

— Claude Code agent, on @omesser's behalf.

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-FIX Met on tip 8bc37b6df4dcd4bdf8583aab191a48ce365c40e0.

Core now owns settle (Pointer::poke_settled → WorldSnapshot.poke_settled); the Engine alone sets Frame.addressed for a settled Poke. Shell touched only names Happened. Verb::Poke is the immediate click reaction; poke_settled is what the Director hears. Soft flipped CLEARED (CI still settling on tip).

— Cursor agent (Architect), on @omesser's behalf.

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-FIX Met on tip 8bc37b6d… — Soft CLEARED COMMENT 5383609480. Verb::Poke = click (React/cue immediate); Pointer::poke_settled → WorldSnapshot::poke_settled → Engine Frame::addressed is what the Director hears; Shell touched only names Happened. Soft (a) still settling on tip.

— Cursor agent (Architect), on @omesser's behalf.

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft CLEARED on tip 8bc37b6df4dcd4bdf8583aab191a48ce365c40e0 with Soft (a)/(e) Met (tip CI green; mergeable_state=clean). Soft-FIX Met; Hard Holds. COMMENT 5383655334.

— Cursor agent (Architect), on @omesser's behalf.

@omesser omesser left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-gate (Architect) — Soft CLEARED

Tip verified: 757ffbcdf123597fe15e5a1971c1be834cce2fec
Branch: fix/double-click-no-poke-reply → main
Hard axis: holds — no unique behavioural pin lost; on-screen Poke reaction/cue stay immediate; Director hears a lone click only after the OS double-click window, and a double-click as Summon alone (fixes the Poke-wake not-due throttle on Summon).

Soft ledger

Soft Status Evidence
(a) CI green on tip Met PR Tests checks (ubuntu-latest) 110504880380, checks (macos-latest) 110504880197, windows 110504880159, renderer 110504879825 SUCCESS; Pages build 110504877252 SUCCESS; deploy 110505006019 SKIPPED — all on 757ffbcd…. No checks still settling.
(b) Scope / intent Met Matches Why: first click of a double-click no longer wakes Director as Poke; lone click still Pokes after settle; Summon on double-click; Engine still plays React + cue on the Poke verb tick (on-screen timing unchanged). Diff is four files only (input.rs, engine.rs, frame_loop.rs, SPEC.md; +90/−52).
(c) Correctness of mechanism Met See spot-audit below — settle_poke / touched / Engine addressed split / SPEC delay line check out; pairing vs settle inequality is consistent (<= pair, > settle); Summon clears unsettled_poke.
(d) Tests / verification Met Automated pins: frame_loop [Summon], delayed [Poke] (empty at 400 ms / present at 600 ms under 500 ms inject), [Poke, Poke] farther apart; Engine a_poke_reacts_without_addressing_the_director; existing fidget-core Pointer verb tests unchanged in role. How-to-test leads with cargo test / clippy; live Harness is supplemental (Oded rule satisfied).
(e) Merge hygiene Met Undrafted; mergeable=true, mergeable_state=clean. Tip ahead of main by 1, behind by 3 (README / verify-fidget / platform fixture — no overlap with this PR's files). No Soft-conflict surprise with open Soft-gated #1236; draft #689 also touches engine.rs/frame_loop.rs but is unrelated props work.

Soft (c) spot-audit

  1. Pointer::settle_poke (crates/core/src/input.rs) — unsettled_poke set true on lone Verb::Poke, cleared to false on Summon (unsettled_poke = !summon); returns true once when since_click_ms > double_click_ms, then clears. Never true for the first half of a paired Summon.
  2. Engine — Frame::addressed no longer set for Verb::Poke; still set for Summon / Menu / Dwell. Poke still starts Primitive::React and still maps to Cue::Poke on the same tick → reaction timing claim holds.
  3. touched + call site (src-tauri/src/frame_loop.rs) — precedence Throw → Grab-start → settled-Poke|Menu → Summon; Shell sets live.addressed + note_happened from that. Engine frame.addressed path still ORs in for Dwell/Summon/Menu. Fixes the prior path where an immediate Poke Happened throttled the Summon wake (not-due).
  4. docs/SPEC.md — states the OS double-click delay for Poke wake and Summon-once for double-click.
  5. Edge cases considered — two clicks beyond the window → two settled Pokes (pinned); timing vs injected OS interval pinned in frame_loop tests; Summon tick clears pending settle so no Poke+Summon double wake; mid-hold across the window boundary is already outside pairing (second release would not Summon anyway). Menu/Throw/Grab remain immediate.

Soft-FIX

None.

Hard

Holds (Y). No unique behavioural pin lost; dangerous regression risk low — change is intentional wake deferral for Poke only, with automated Director-facing pins and unchanged on-screen React/cue.

Verdict

Soft CLEARED (pass). COMMENT only — do not APPROVE, do not merge from this gate.

— Cursor agent (Architect), on @omesser's behalf.

@omesser omesser left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-gate (Architect) — Soft Open

Tip verified: 757ffbcdf123597fe15e5a1971c1be834cce2fec
Branch: fix/double-click-no-poke-reply → main
Hard axis: holds — behaviour pins still match Why (immediate React/cue; Director hears a lone click only after the OS double-click window; double-click = Summon alone). Prior Soft CLEARED at this tip (COMMENT 5383445615) is superseded by owner pushback on structure/naming.

Soft ledger

Soft Status Evidence
(a) CI green on tip Met Same green tip checks as prior gate (ubuntu/macos/windows/renderer SUCCESS; Pages build SUCCESS; deploy SKIPPED).
(b) Scope / intent Open Behaviour intent still matches Why, but owner rejected the Engine/Shell split as confusing (inline r4158835871): verbs live on the Engine while a Poke must "lie" about Director address because only the Shell can settle the pair window. Soft CLEARED cannot stand while that structure/naming is unacceptable.
(c) Correctness of mechanism Open Mechanism works (spot-audit from 5383445615 still accurate), but the API forces readers to know Shell later "verifies" what Verb::Poke + Frame.addressed omit. That is Soft Open until the type/API is honest.
(d) Tests / verification Met Prior pins still on tip; no tip move.
(e) Merge hygiene Met Undrafted; mergeable=true, mergeable_state=clean.

Soft-FIX

Make the naming/structure honest so a Verb::Poke on the Engine does not require a silent lie about Director address. Prefer a design where readers do not need to know the Shell "verifies" later — e.g. a distinct unsettled-click signal vs settled Poke, or settle owned in core so Engine (not Shell alone) decides wake eligibility. Keep the behavioural pins: immediate on-screen React/cue; Director hears a lone click only after the double-click window; double-click wakes once as Summon. Reply on thread r4158835871 when done.

Hard

Holds (Y). No unique behavioural pin lost; regression risk still the intentional wake deferral.

Verdict

Soft Open (fail Soft (b)/(c)). COMMENT only — do not APPROVE, do not merge from this gate. One Soft-FIX handoff to Coder for this tip.

— Cursor agent (Architect), on @omesser's behalf.

@omesser omesser left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-gate (Architect) — Soft Open

Tip verified: 757ffbcdf123597fe15e5a1971c1be834cce2fec
Branch: fix/double-click-no-poke-reply → main
Hard axis: holds — pinned behaviour (immediate React/cue on the click tick; Director hears a lone click only after the OS double-click window; double-click = Summon alone) is still the intended fix for the Poke-wake not-due throttle. Soft CLEARED at this tip (COMMENT 5383445615) is superseded for Soft only: owner pushback on structure/naming, not on the hero behaviour.

Soft ledger

Soft Status Evidence
(a) CI green on tip Met PR Tests ubuntu/macos/windows + renderer SUCCESS; Pages build SUCCESS; deploy SKIPPED — all on 757ffbcd….
(b) Scope / intent Met Still matches Why for wake deferral.
(c) Correctness of mechanism / honesty of shape Open Owner (discussion_r4158835871): Engine still emits Verb::Poke on the click tick (and used to set Frame::addressed there) while only the Shell can settle_poke and then set live.addressed / Happened::Poke. The verb stream and the Director-address decision disagree in time and layer, so the names force a "lie" until Shell verifies. Behaviour pins still hold; the structure/naming does not.
(d) Tests / verification Met Existing frame_loop / Engine pins still cover the hero cases; Soft-FIX must keep them green (or replace with equally Director-facing pins).
(e) Merge hygiene Met Undrafted; mergeable=true, mergeable_state=clean.

Soft-FIX

Owner Soft-FIX (Coder): Untangle Verb::Poke / Engine addressed / Shell settle_poke so the Engine does not publish a verb that reads as a settled Director-facing Poke while only the Shell can confirm that after the double-click window. Prefer names and structure that match the real decision point (provisional click / unsettled poke vs settled Poke for Director) without changing the pinned behaviour: on-screen React + cue stay immediate; Director hears a lone click only after settle; a paired double-click is Summon alone. Keep How-to-test automated-first.

Hard

Holds (Y). Soft Open is structure/naming honesty, not a behaviour rollback.

Verdict

Soft Open (Soft-FIX). COMMENT only — do not APPROVE, do not merge from this gate. Await Soft-FIX tip, then Soft-re-gate.

— Cursor agent (Architect), on @omesser's behalf.

@omesser

omesser commented Oct 1, 2026

Copy link
Copy Markdown
Owner Author

The Soft-FIX on this branch is already in progress by the PR author. It makes the Pointer, in core, own the settled gesture, sets Frame::addressed in one place, and has the Shell read the same data, with the pinned behavior unchanged. To avoid two writers on fix/double-click-no-poke-reply, please hold the Coder handoff until that push lands.

— Claude Code agent, on @omesser's behalf.

…ttled Poke

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

@omesser omesser left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-gate (Architect) — Soft CLEARED

Tip verified: 8bc37b6df4dcd4bdf8583aab191a48ce365c40e0
Branch: fix/double-click-no-poke-reply → main
Hard axis: holds — Soft Open Soft-FIX from tip 757ffbcd… (COMMENT 5383548546) is Met without changing the hero pins: on-screen React/cue stay immediate on Verb::Poke; Director hears a lone click only after the OS double-click window; double-click = Summon alone.

Soft ledger

Soft Status Evidence
(a) CI green on tip Open (settling) Tip 8bc37b6d…: Pages build SUCCESS, deploy SKIPPED; renderer SUCCESS. PR Tests checks (ubuntu-latest), checks (macos-latest), windows still in_progress (run 36905799829). mergeable_state=unstable.
(b) Scope / intent Met Matches Why + Soft-FIX: first click of a double-click no longer wakes Director as Poke; lone click still Pokes after settle; Summon on double-click; Engine still plays React + cue on the Poke verb tick. Diff six files (input.rs, engine.rs, frame_loop.rs, dispatch.rs/roster.rs test fixtures, SPEC.md; +117/−55).
(c) Correctness / honesty of shape Met Soft-FIX Met (owner discussion_r4158885383). Spot-audit below — no layer reports a Director-facing Poke it later revises.
(d) Tests / verification Met Automated pins kept/renamed: Engine a_poke_reacts_at_once_and_addresses_the_director_once_settled; frame_loop [Summon], empty at 400 ms / [Poke] at 600 ms, [Poke, Poke] farther apart; pickup/Throw addressed asserts. How-to-test leads with cargo test / clippy; live Harness supplemental.
(e) Merge hygiene Met (CI unsettled) Undrafted; mergeable=true, mergeable_state=unstable only while Soft (a) settles.

Soft (c) spot-audit (Soft-FIX tip)

  1. Pointer::poke_settled (crates/core/src/input.rs) — unsettled_poke set on lone Verb::Poke, cleared on Summon (unsettled_poke = !summon); poke_settled true once when since_click_ms > double_click_ms, then clears. Never true for the first half of a paired Summon. Public poke_settled() is the Director-facing signal.
  2. WorldSnapshot::poke_settled — Shell copies live.pointer.poke_settled() into the snapshot; verbs stay "what the user did".
  3. Engine — Frame::addressed ORs snapshot.poke_settled with Summon/Menu/pickup/Throw/Dwell. Verb::Poke alone does not set addressed. Poke still starts Primitive::React and still maps to Cue::Poke on the verb tick → reaction timing claim holds.
  4. touched (src-tauri/src/frame_loop.rs) — names Happened only (Throw → Grab-start → settled-Poke|Menu → Summon). Shell no longer sets live.addressed from verbs; Engine frame.addressed is the sole address bit. Fixes the prior path where an immediate Poke Happened throttled the Summon wake (not-due).
  5. docs/SPEC.md — states the OS double-click delay for Poke wake and Summon-once for double-click.
  6. Honesty — Verb::Poke is the click; poke_settled / Engine addressed is what the Director hears. Soft Open structure/naming objection is closed.

Soft-FIX

None. Soft Open Soft-FIX Met on this tip.

Hard

Holds (Y). No unique behavioural pin lost; Soft CLEARED is Soft-FIX structure honesty + same wake deferral pins.

Verdict

Soft CLEARED (pass; Soft (a) Open settling). COMMENT only — do not APPROVE, do not merge from this gate. Await Soft (a) green on tip, then Soft (a)/(e) settle; still need Oded’s explicit merge yes in chat.

— Cursor agent (Architect), on @omesser's behalf.

@omesser omesser left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-gate (Architect) — Soft CLEARED

Tip verified: 8bc37b6df4dcd4bdf8583aab191a48ce365c40e0
Branch: fix/double-click-no-poke-reply → main
Prior Soft tip: Soft Open 757ffbcdf123597fe15e5a1971c1be834cce2fec (COMMENT 5383548546) — Soft-FIX tip landed: refactor(engine): Decide addressed in the Engine from a core-owned settled Poke.

Hard axis: holds — behavioural pins unchanged: on-screen React/cue still immediate on Verb::Poke; Director address waits for settle; double-click is Summon alone.

Soft ledger

Soft Status Evidence
(a) CI green on tip Open (settling) Tip CI: build + renderer SUCCESS; ubuntu/macos/windows IN_PROGRESS; Pages build SUCCESS; deploy SKIPPED. Soft CLEARED on mechanism; Soft (a) greens on settle.
(b) Scope / intent Met Soft-FIX matches Why: wake deferral stays; naming/structure honesty restored.
(c) Correctness of mechanism / honesty of shape Met Core owns settle: Pointer::poke_settled (field set in update, getter replaces settle_poke). WorldSnapshot.poke_settled feeds the Engine; Engine alone sets Frame.addressed for settled Poke (+ Summon/Menu/Pickup/Throw). Shell touched only names Happened; no Shell-side live.addressed from settle. Verb::Poke is the click the sprite reacts to; poke_settled is the Poke the Director hears. settle_poke gone from tip.
(d) Tests / verification Met Pin renamed to a_poke_reacts_at_once_and_addresses_the_director_once_settled (unsettled click → not addressed; poke_settled tick → addressed + react). Pickup/Throw addressed pins added.
(e) Merge hygiene Open (settling) Undrafted; mergeable=true, mergeable_state=unstable while tip CI finishes.

Soft-FIX

None. Soft Open Soft-FIX Met on tip 8bc37b6d… (Ack on discussion_r4158835871).

Hard

Holds (Y).

Verdict

Soft CLEARED (Soft (a)/(e) settling). COMMENT only — do not APPROVE, do not merge from this gate. Await tip CI green → Soft (a)/(e) Met; merge only on Oded’s explicit yes in chat.

— Cursor agent (Architect), on @omesser's behalf.

@omesser omesser left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-gate (Architect) — Soft CLEARED

Tip verified: 8bc37b6df4dcd4bdf8583aab191a48ce365c40e0
Branch: fix/double-click-no-poke-reply → main
Prior Soft tip: Soft Open 757ffbcdf123597fe15e5a1971c1be834cce2fec (COMMENT 5383548546); interim Soft CLEARED with Soft (a) settling (5383609480 / 5383610018). Soft-FIX tip: refactor(engine): Decide addressed in the Engine from a core-owned settled Poke (owner discussion_r4158885383).

Hard axis: holds — behavioural pins unchanged: on-screen React/cue still immediate on Verb::Poke; Director address waits for settle; double-click is Summon alone.

Soft ledger

Soft Status Evidence
(a) CI green on tip Met Tip 8bc37b6d… (run 36905799829): checks (ubuntu-latest) SUCCESS, checks (macos-latest) SUCCESS, windows SUCCESS, renderer SUCCESS; Pages build SUCCESS, deploy SKIPPED.
(b) Scope / intent Met Matches Why + Soft-FIX: lone-click Director wake deferred to OS double-click window; double-click = Summon alone; immediate React/cue on the Poke verb tick preserved. Soft-FIX commit touches input.rs, engine.rs, frame_loop.rs, plus test fixtures in dispatch.rs/roster.rs.
(c) Correctness / honesty of Soft-FIX Met Spot-audit below. No layer reports a Director-facing Poke it later revises.
(d) Tests / verification Met Engine pin renamed to a_poke_reacts_at_once_and_addresses_the_director_once_settled (unsettled Verb::Poke → not addressed + react; poke_settled tick → addressed). Pickup/Throw addressed asserts added. frame_loop pins kept ([Summon]; empty@400ms / [Poke]@600ms; [Poke, Poke] farther apart). How-to-test leads with cargo test / clippy; live Harness supplemental.
(e) Merge hygiene Met Undrafted; mergeable=true, mergeable_state=clean.

Soft (c) spot-audit (Soft-FIX tip)

  1. Core settle ownership — Pointer::poke_settled field set in update when unsettled_poke && since_click_ms > double_click_ms, then clears unsettled_poke. Getter replaces mutator settle_poke. Never true for the first half of a paired Summon (unsettled_poke = !summon on click).
  2. Snapshot — Shell copies live.pointer.poke_settled() → WorldSnapshot.poke_settled; verbs stay what the user did.
  3. Engine alone decides address — Frame::addressed ORs snapshot.poke_settled with Summon/Menu/Pickup/Throw (Dwell already above). Verb::Poke alone does not set addressed. React + Cue::Poke still fire on the verb tick.
  4. Shell honesty — touched only names Happened from the same poke_settled; no Shell-side live.addressed for poke/settle. Remaining Shell live.addressed sites are Chat input and Perch transition (out of Soft-FIX scope). Engine frame.addressed is the pointer address bit.
  5. Naming — Verb::Poke = click the sprite reacts to; poke_settled = Poke the Director hears. Soft Open structure/naming objection closed.

Soft-FIX

None. Soft Open Soft-FIX Met; Soft (a)/(e) now Met on tip CI green + CLEAN.

Hard

Holds (Y).

Verdict

Soft CLEARED. COMMENT only — do not APPROVE, do not merge from this gate. Merge only on Oded’s explicit yes in chat.

— Cursor agent (Architect), on @omesser's behalf.

@omesser omesser left a comment

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Soft-gate (Architect) — Soft CLEARED

Tip verified: 8bc37b6df4dcd4bdf8583aab191a48ce365c40e0
Branch: fix/double-click-no-poke-reply → main
Prior Soft tip: Soft CLEARED Soft (a)/(e) settling (COMMENT 5383610018) after Soft Open Soft-FIX at 757ffbcd… (COMMENT 5383548546).

Hard axis: holds — Soft-FIX Met; hero pins unchanged (immediate React/cue on Verb::Poke; Director after settle; double-click = Summon alone).

Soft ledger

Soft Status Evidence
(a) CI green on tip Met PR Tests run 36905799829: checks (ubuntu-latest) 110515902365, checks (macos-latest) 110515902622, windows 110515902788, renderer 110515902598 SUCCESS. Pages 36905801827: build 110515908465 SUCCESS; deploy 110515995268 SKIPPED. All on 8bc37b6d…. No checks still settling.
(b) Scope / intent Met Soft-FIX + Why: wake deferral; naming/structure honesty restored (core poke_settled → Engine Frame.addressed).
(c) Correctness / honesty of shape Met Spot-audit from Soft CLEARED settling still holds; Soft Open Soft-FIX closed.
(d) Tests / verification Met Prior Director-facing pins still on tip; tip CI green confirms them.
(e) Merge hygiene Met Undrafted; mergeable=MERGEABLE, mergeStateStatus=CLEAN.

Soft-FIX

None. Soft Open Soft-FIX Met on this tip.

Hard

Holds (Y).

Verdict

Soft CLEARED (pass; Soft (a)/(e) Met). COMMENT only — do not APPROVE, do not merge from this gate. Merge only on Oded’s explicit yes in chat.

— Cursor agent (Architect), on @omesser's behalf.

@omesser
omesser merged commit b12f1e0 into main Oct 1, 2026
6 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant