feat(port): adopt the measured held-direction repeat rate (F1)

REPEAT_DELAY = 0.402, REPEAT_INTERVAL = 0.134, from
docs/re/f1-repeat-measured-via-driver-patch.md -- 12 and 4 frames at the
run's achieved 29.87 fps guest rate, converted to seconds because this port
does not run at the guest's rate and it is the cadence that was measured.

The mechanism has been here since 2026-09-02 and inert on purpose. The
instruction it was waiting on is now vindicated in the most awkward way: the
draft it refused to ship had 0.40 / 0.20, so the guessed delay was nearly
right and the guessed interval was off by 50 %. The half that was wrong
would have been protected by the half that was right.

The delay and the interval are NOT equally well evidenced, and the code says
so at the constants. No physical controller exists in the Decoder's
container, so the measurement fed Canary's file driver the SDL driver's own
400/100 ms constants: the 402 ms that came back is the constant that went
in, and confirms the instrument. The 133 ms interval against a fed-in 100 ms
is the new fact -- the game paces repeats to its own frame consumption. One
run; the two-run minimum is not met and the finding says so itself.

Also: the prediction that this would turn verify-input's "a held stick is
ONE step, not six" red was wrong. It stayed green, because steps() never
advances a clock and so had never called repeat_due() at all -- the rate was
about to ship into a harness with no coverage of the feature, with a green
line that would have been read as coverage.

So verify-input gains a `repeat` subject: nothing before the delay, the
first repeat on the delay, the steady interval, cadence independent of frame
rate (the code claims this in a comment, so it is now asserted), and a
direction change restarting the delay. Each asserts THESE numbers, not the
shape -- a shape-only check would have passed on 0.40 / 0.20. The control
removes the premise, a held direction, and every controllable row inverts.

The first-repeat row measures from the arming tick, not from t=0: that tick
is the frame the press is handled, which is the origin the finding measures
its 12 frames from. Measured from zero it read 0.433 vs 0.402 and the
tolerance would have had to be widened to hide a units mismatch.

Closes #2

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

docs/port/held-direction-repeat.md

@@ -1,7 +1,10 @@
-# F1 — the menu repeats on a held direction: mechanism shipped, **rate deliberately not**
+# F1 — the menu repeats on a held direction: mechanism shipped, **and the rate adopted**
 
-**Status:** ✅ mechanism implemented and wired. 🔴 **inert on purpose** — it does
+**Status:** ✅ mechanism implemented and wired, 2026-09-02. ✅ **rate adopted
-nothing until a measured repeat rate exists. Written 2026-09-02 by the Port.
+2026-09-13** from `docs/re/f1-repeat-measured-via-driver-patch.md` —
+`REPEAT_DELAY = 0.402`, `REPEAT_INTERVAL = 0.134`. It ran inert for eleven days
+and that was the right state; this page keeps the inert-era reasoning because it
+is why the two numbers can be trusted now.
 
 ## What was reported
 
@@ -57,42 +60,87 @@ A repeat that could fire during a movie or mid-transition would be a **second,
 subtly different input path**, and the first thing this port learned about input
 is that a second path is where the defect hides.
 
-## 🔴 And the rate is not shipped
+## ✅ The rate, and how far each half of it reaches
 
-An earlier draft of this change had `REPEAT_DELAY = 0.40` and
+An earlier draft had `REPEAT_DELAY = 0.40` and `REPEAT_INTERVAL = 0.20` with a
-`REPEAT_INTERVAL = 0.20`, with a paragraph explaining that they were authored.
+paragraph explaining that they were authored. They were removed rather than
-**They were removed rather than commented out**, on an explicit instruction:
+commented out, on an explicit instruction:
 
 > *"Take the RATE from the Decoder — do NOT ship a placeholder interval. An
 > invented rate here is indistinguishable from a measured one later, and this is
 > the exact field where that already cost us."*
 
-The instruction is right and the draft was the named failure mode: the
+📌 **The measurement has now landed, and it vindicates the instruction in the
-explanation would have merged, the numbers would have felt roughly right, and
+most awkward possible way: the guessed delay was nearly right and the guessed
-nothing downstream could have separated them from a measurement. `REPEAT_DELAY`
+interval was off by 50 %.** 0.40 against a measured 0.402; 0.20 against a
-is `-1.0`; `repeat_due()` returns 0 while `repeat_rate_known()` is false.
+measured 0.134. Had both shipped, the half that was wrong would have been
+protected by the half that was right.
 
-**One thing about the rate IS measured, and it narrows the question.** The game
+The Decoder measured, in Canary at an achieved 29.87 fps guest rate, **12 frames**
-digitises the left stick to four direction bits at 61 % deflection, so it cannot
+from the press-triggered step to the first repeat and **4 frames** per step after
-see deflection magnitude at all — the repeat it drives *cannot* be
+it. Converted to seconds here because this port does not run at the guest's rate
-faster-the-harder-you-push. That excludes the one competing model, so only two
+and it is the cadence that was measured: 12 / 29.87 = **0.402**, 4 / 29.87 =
-constants are open and a single measurement closes both.
+**0.134**.
 
-## ⚠️ Adopting the rate breaks a green check, for the right reason
+### ⚠️ The two numbers are not equally well evidenced
 
-`tools/port/verify-input` asserts *"a held stick is ONE step, not six"*. That row
+No physical controller exists in the Decoder's container. The measurement was
-passes today **because the feature is inert**, i.e. it asserts the absence of the
+taken by patching Canary's `--hid=file` driver to emit Keystroke `REPEAT` at the
-repeat. When a rate is adopted a held stick should produce further steps and that
+**SDL driver's own** 400 ms / 100 ms constants:
-row will go red.
 
-It is not wrong and it should not be deleted in a hurry: it was written for the
+* the **delay** came back as 402 ms — to within the frame quantum, *the constant
-2026-09-01 jitter defect, so it will *look* like that bug returning. It has to be
+  that was fed in*. It confirms the instrument, not the game;
-re-stated as "one step per deflection **plus** the measured repeat", with the
+* the **interval** came back as 133 ms against a fed-in 100 ms. That gap is the
-jitter case still covered inside the delay window.
+  genuinely new fact: the game paces repeats to its own frame consumption rather
+  than to the event stream.
+
+So the interval is what the game does; the delay is what Xenia's SDL driver does,
+and the game was not observed to disagree. **One run** — the corpus's own two-run
+minimum is not met, and the source page says so itself.
+
+**One thing about the rate was already measured, and it narrowed the question.**
+The game digitises the left stick to four direction bits at 61 % deflection, so
+it cannot see deflection magnitude at all — the repeat it drives *cannot* be
+faster-the-harder-you-push. That excluded the one competing model, so only two
+constants were ever open and a single measurement closed both.
+
+## 🔴 Adopting the rate was predicted to break a green check. It did not, and that was worse
+
+This page and `gamepad.gd` both said that `verify-input`'s row *"a held stick is
+ONE step, not six"* asserts the **absence** of the repeat, and would go red on
+adoption — looking like the 2026-09-01 jitter defect returning.
+
+**Run on adoption day: the row stayed green.** `steps()` feeds axis values through
+the latch and never advances a clock, so it had never called `repeat_due()` at
+all. The row tests the *latch*, which the repeat does not touch. The prediction
+was reasoned rather than run.
+
+What it hid is the real problem: the rate was about to ship into a harness with
+**no coverage of this feature whatsoever**, and that green line would have been
+read as coverage of it.
+
+The fix was not to change that row. `verify-input` gains a `repeat` subject that
+holds a direction through the same latch and ticks `repeat_due()`, asserting
+**these two numbers** — a shape-only check ("it repeats eventually") would have
+passed on the 0.40 / 0.20 guess this port refused to ship. Five rows: nothing
+before the delay, the first repeat on the delay, the steady interval, cadence
+independent of frame rate (the code claims this in a comment, so it is asserted),
+and a direction change restarting the delay.
+
+⚠️ **The origin is one frame, and it is not a tolerance.** `repeat_due()`'s first
+call only latches the direction; the clock accumulates from the call after it. In
+the port that first call is the frame the press is handled — the frame that
+produced the press-triggered step — which is the origin the finding measures its
+12 frames from. Measured from `t = 0` the harness read 0.433 against 0.402 and
+the tolerance would have had to be widened to hide a units mismatch.
 
 ## What this does not claim
 
-* That the repeat feels right. It cannot — it does not run.
+* That the repeat feels right. It runs now, and **only a human can answer that**:
+  0.134 s is ~7.5 steps a second, and the play-test that opened this asked for
+  *"slow enough to see which item is selected"*. If it reads as too fast, the
+  100 ms constant is Xenia's and not the game's, and no capture in that
+  container would have revealed it.
 * Any rate, or any bound on one. "Medium pace" is a direction, not a number, and
   it is not recorded anywhere as data.
 * That the d-pad and the stick repeat at the *same* rate. Both repeat; nobody

port/scripts/gamepad.gd

@@ -100,17 +100,43 @@ const RELEASE := 0.4
 ## too."* So the FACT covers both input devices, which is why `held_direction()`
 ## polls the pad and the keyboard and not just the stick.
 ##
-## 🔴 **THE RATE IS DELIBERATELY UNSET, AND THE REPEAT DOES NOT RUN UNTIL IT IS
+## ✅ **THE RATE IS NOW MEASURED, 2026-09-12, and adopted here.**
-## MEASURED.** The instruction is explicit: *"Take the RATE from the Decoder — do
+## `docs/re/f1-repeat-measured-via-driver-patch.md`, with the per-transition
-## NOT ship a placeholder interval. An invented rate here is indistinguishable
+## reference data in `docs/re/data/f1-repeat-cursor-transitions.tsv`.
-## from a measured one later, and this is the exact field where that already cost
-## us."*
 ##
-## An earlier draft of this file had 0.40 / 0.20 with a paragraph explaining that
+## ~~THE RATE IS DELIBERATELY UNSET, AND THE REPEAT DOES NOT RUN UNTIL IT IS
-## they were authored. **That is precisely the failure mode named above** — the
+## MEASURED.~~ It ran unset for eleven days and that was the right state; the
-## explanation would have been merged, the numbers would have felt roughly right,
+## paragraph is struck through rather than deleted because the reason it gave is
-## and nothing afterwards could distinguish them from a measurement. They are
+## the reason these two numbers can be trusted now. An earlier draft had
-## removed rather than commented out.
+## **0.40 / 0.20** with a note saying they were authored — and 0.40 would have
+## looked vindicated today while 0.20 was off by 50 %. That is exactly why an
+## explained guess is worse than none: half of it would have been right.
+##
+## ## Where the two numbers come from, and how far they reach
+##
+## Measured in Canary at an achieved **29.87 fps** guest rate: **12 frames**
+## from the press-triggered step to the first repeat, then **4 frames** per step
+## (13 of 15 gaps; 3 frames for the other 2). Converted to seconds here, not
+## frames, because this port does not run at the guest's rate and it is the
+## *cadence* that was measured — 12 / 29.87 = 0.402, 4 / 29.87 = 0.134.
+##
+## ⚠️ **THE DELAY IS WEAKER EVIDENCE THAN THE INTERVAL, and they should not be
+## trusted equally.** No physical controller exists in that container, so the
+## measurement was taken by patching Canary's `--hid=file` driver to emit
+## Keystroke `REPEAT` at the SDL driver's own 400 ms / 100 ms constants. The
+## 402 ms that came back is, to within the frame quantum, **the constant that
+## was fed in** — it confirms the instrument, not the game. The 133 ms interval
+## is the genuinely new fact: the driver was fed 100 ms and the cursor moved
+## every 133, so the game paces repeats to its own frame consumption rather
+## than to the event stream.
+##
+## So: the interval is what the game does. The delay is what Xenia's SDL driver
+## does, and the game was not observed to disagree with it. If a capture through
+## a real controller ever contradicts 0.402, that is the number to move.
+##
+## 📌 One run. The corpus's own two-run minimum is **not met** — the source page
+## says so itself, and this comment repeats it rather than letting the constant
+## look firmer at the call site than it does at the finding.
 ##
 ## 📌 **A constant interval is the right SHAPE, and that part IS measured.** The
 ## game digitises the left stick to four direction bits at 61 % deflection
@@ -118,20 +144,27 @@ const RELEASE := 0.4
 ## drives cannot be faster-the-harder-you-push. That excludes the one competing
 ## model, so only the two constants are open, and one measurement closes both.
 ##
-## ⚠️ **TO ADOPT, TWO THINGS CHANGE, NOT ONE.** Set both constants to the
+## ⚠️ ~~**TO ADOPT, TWO THINGS CHANGE, NOT ONE.** … that green row would go red
-## measured seconds — and update `tools/port/verify-input`, whose row *"a held
+## for the right reason and be read as a regression.~~
-## stick is ONE step, not six"* currently asserts **the absence of this
-## feature**. It passes today because the repeat is inert; the moment a rate is
-## adopted a held stick SHOULD produce further steps, and that green row would
-## go red for the right reason and be read as a regression.
 ##
-## 📌 That row is not wrong. A check written against today's behaviour becomes an
+## 🔴 **RUN ON ADOPTION DAY: THE ROW STAYED GREEN, AND THAT IS WORSE.**
-## assertion that the behaviour never changes, and this one has the additional
+## `verify-input`'s `steps()` never advances a clock, so it had never called
-## trap of looking like a bug-fix regression test — it was written for the
+## `repeat_due()` at all — the row it warned about tests the *latch*, which the
-## jitter defect, and the repeat is not that defect returning.
+## repeat does not touch. The prediction was reasoned rather than run, and what
-## `pad-repeat` in `BLOCKED.md` carries the request.
+## it hid is the real problem: the rate was about to ship into a harness with
-const REPEAT_DELAY := -1.0
+## **no coverage of this feature whatsoever**, and the green line would have
-const REPEAT_INTERVAL := -1.0
+## been read as coverage.
+##
+## The fix was not to change that row. It was to add a `repeat` subject that
+## holds a direction through the same latch and ticks `repeat_due()`, asserting
+## **these two numbers** rather than "it repeats eventually" — a shape-only
+## check would have passed on the 0.40 / 0.20 guess this file refused to ship.
+##
+## 📌 The original point survives intact and is worth keeping: a check written
+## against today's behaviour becomes an assertion that the behaviour never
+## changes. It was simply aimed at the wrong row.
+const REPEAT_DELAY := 0.402
+const REPEAT_INTERVAL := 0.134
 
 
 ## Whether a measured repeat rate has been adopted. Until it has, the port keeps

tools/port/verify-input

@@ -112,6 +112,53 @@ func steps(values: Array, latched: bool) -> int:
 func nav(values: Array) -> int:
 	return steps(values, mode != "control")
 
+## Hold a direction through the REAL latch, then tick `repeat_due()` and report
+## the time of every repeat it produces.
+##
+## The press edge is fed through `accepts()` rather than poked into the latch,
+## so this exercises the same path the port does -- `held_direction()` reads
+## that latch first. Under `--control` the hold is simply not made: the rate is
+## a `const` and cannot be removed at runtime, so what the control removes is
+## the PREMISE (a direction being held), and every count must go to zero.
+func hold_and_tick(seconds: float, delta: float) -> Array[float]:
+	var pad := Gamepad.new()
+	if mode != "control":
+		pad.accepts(deflect(0.92))
+	var t := 0.0
+	var out: Array[float] = []
+	while t < seconds:
+		t += delta
+		if pad.repeat_due(delta) != 0:
+			out.append(t)
+	return out
+
+## Hold one way past the delay, reverse, and report how long until the first
+## repeat in the NEW direction. Inheriting the old cadence would show up here as
+## a time far below `REPEAT_DELAY`.
+func reversal_delay(delta: float) -> float:
+	var pad := Gamepad.new()
+	if mode != "control":
+		pad.accepts(deflect(0.92))
+	var t := 0.0
+	while t < 1.0:
+		t += delta
+		pad.repeat_due(delta)
+	pad.accepts(deflect(0.0))
+	if mode != "control":
+		pad.accepts(deflect(-0.92))
+	t = 0.0
+	while t < 2.0:
+		t += delta
+		if pad.repeat_due(delta) != 0:
+			return t
+	return -1.0
+
+func deflect(v: float) -> InputEventJoypadMotion:
+	var e := InputEventJoypadMotion.new()
+	e.axis = JOY_AXIS_LEFT_Y
+	e.axis_value = v
+	return e
+
 func _init() -> void:
 	# The control removes the repair. Everything else runs with it applied.
 	if mode != "control":
@@ -201,6 +248,76 @@ func _init() -> void:
 			passed += 1
 	ok("d-pad presses are not latched", "negative", passed == 3, "%d of 3" % passed)
 
+	# ── 4. subject `repeat` -- a held direction repeats at the MEASURED rate ──
+	#
+	# 🔴 THIS SECTION EXISTS BECAUSE A PREDICTION IN `gamepad.gd` WAS WRONG.
+	# That file said adopting the rate would turn "a held stick is ONE step, not
+	# six" red, and warned that the row would read as a regression. Measured on
+	# adoption day: it stays green, because `steps()` never advances a clock and
+	# so has never called `repeat_due()` at all. The warning was reasoned, not
+	# run -- and the real consequence is worse than the one predicted. The rate
+	# shipped into a harness with **no coverage of the feature whatsoever**.
+	#
+	# The rows below are that coverage. They assert the two numbers from
+	# `docs/re/f1-repeat-measured-via-driver-patch.md`, not the shape alone: a
+	# test that only checked "it repeats eventually" would pass on any constant
+	# and would have passed on the 0.40 / 0.20 guess this port deliberately
+	# refused to ship.
+	var FRAME := 1.0 / 60.0
+
+	# Hold the stick by pushing it through the same latch the port uses, then
+	# tick. `held_direction()` reads the latch, so this is the real path.
+	var timeline := hold_and_tick(2.0, FRAME)
+	# ⚠️ MEASURED FROM THE ARMING TICK, NOT FROM t=0, and the difference is a
+	# whole frame. `repeat_due()`'s first call only latches the direction and
+	# returns 0; the clock accumulates from the call after it. In the port that
+	# first call happens on the frame the press is handled -- the frame that
+	# produced the press-triggered step -- and the finding measures its 12
+	# frames "from the press-triggered step to the first repeat". So the arming
+	# tick is the press step, and subtracting it is what puts the harness and
+	# the finding on the same origin. Without this the row read 0.433 vs 0.402
+	# and the tolerance would have had to be widened to hide a units mismatch.
+	var first: float = (timeline[0] - FRAME) if not timeline.is_empty() else -1.0
+
+	ok("nothing repeats before the measured delay", "negative",
+		first >= Gamepad.REPEAT_DELAY,
+		"first repeat %.3fs after the press step, delay is %.3f"
+			% [first, Gamepad.REPEAT_DELAY],
+		"the steady-interval row below")
+	ok("first repeat lands on the measured delay", "repeat",
+		first >= 0.0 and absf(first - Gamepad.REPEAT_DELAY) <= FRAME,
+		"%.3fs vs %.3f (±one frame)" % [first, Gamepad.REPEAT_DELAY])
+
+	var gaps: Array[float] = []
+	for i in range(1, timeline.size()):
+		gaps.append(timeline[i] - timeline[i - 1])
+	var mean := 0.0
+	for g in gaps:
+		mean += g
+	mean = mean / gaps.size() if not gaps.is_empty() else -1.0
+	ok("steady interval is the measured 0.134s", "repeat",
+		not gaps.is_empty() and absf(mean - Gamepad.REPEAT_INTERVAL) <= FRAME,
+		"mean %.3fs over %d gap(s) vs %.3f" % [mean, gaps.size(), Gamepad.REPEAT_INTERVAL])
+
+	# `repeat_due()` subtracts the interval rather than resetting the clock,
+	# with the stated reason "at 140 fps and at 30 fps the same number of steps
+	# happen per second". That is a claim about the code, so it is asserted
+	# rather than believed.
+	var at30 := hold_and_tick(2.0, 1.0 / 30.0).size()
+	var at240 := hold_and_tick(2.0, 1.0 / 240.0).size()
+	# `at30 > 0` matters: with nothing held both counts are 0 and "they agree"
+	# would be a green line for a mechanism that never ran -- the control caught
+	# exactly that, so the count is asserted as well as the agreement.
+	ok("the cadence does not drift with frame rate", "repeat",
+		at30 > 0 and absf(at30 - at240) <= 1,
+		"%d steps at 30fps, %d at 240fps" % [at30, at240])
+
+	# A direction change must restart the delay, not inherit the old cadence --
+	# otherwise flicking the other way mid-repeat steps instantly.
+	ok("a direction change restarts the delay", "repeat",
+		reversal_delay(FRAME) >= Gamepad.REPEAT_DELAY,
+		"%.3fs after the reversal" % reversal_delay(FRAME))
+
 	if ran == 0:
 		print("🔴 no check ran -- the harness asserted nothing")
 		quit(2)