Let the page buttons keep up with the finger: Next, Next, Next reaches page 4 (0.2.72)

The Previous and Next bars ran their tap through the tile guard, whose 600 ms
same-tile window (so one contact never fires a tile twice) dropped the second
and third tap of a quick Next, Next, Next. They now use accept_repeat, the
-/+ keys' rule since 0.2.25: every clean tap counts, only a contact within
150 ms of the last accepted tap on the same button is a bounce. The page fill
already dropped a fill under way when the next page was asked for. Firmware
0.2.61; test_cyd_ui.cpp covers the page buttons; host harness and check builds
in docs/TEST_RESULTS_0272.md.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

components/smart_display/cyd_ui.h

@@ -73,8 +73,9 @@ class TouchGuard {
     if (moved_) { reject_ = MOVED; return false; }
     return accept_slider(now, tile);
   }
-  // For -/+ keys: every clean tap counts, even the third within a second, so a
+  // For -/+ keys and the page buttons: every clean tap counts, even the third within a second,
-  // setpoint moves several steps in one go. Only bounce (same key within `gap`) is dropped.
+  // so a setpoint moves several steps in one go and Next, Next, Next reaches page 4 without
+  // waiting for each page to draw. Only bounce (same key within `gap`) is dropped.
   bool accept_repeat(uint32_t now, int tile, uint32_t gap = 150) {
     if (moved_) { reject_ = MOVED; return false; }
     if (accepted_) { reject_ = USED; return false; }

docs/RELEASING.md

@@ -176,6 +176,21 @@ icons sit off-center in the browser); run `generate_packages.py` afterward.
 The `MDI_GLYPH_*` substitutions are retired: `TILEn_ICON` in manual
 profiles must come from the set. No changed preferences or keys.
 
+### Compatibility 0.2.72 / firmware 0.2.61
+
+Firmware only (both board profiles; `cyd_ui.h` gains a comment); the app raises `FIRMWARE_VERSION`. No protocol,
+storage or editor change, and an older app drives this firmware as before.
+
+- The `< Previous` and `Next >` bars on `home_page` call `cyd::touch_guard.accept_repeat(millis(), 11 / 12)` instead
+  of `accept()`. `accept()` keeps a 600 ms window per tile so one contact can never fire a tile's action twice; on the
+  page bar that window dropped the second and third tap of Next, Next, Next, so page 4 took three waits of about
+  half a second. `accept_repeat`, the -/+ keys' rule since 0.2.25, still drops a contact shorter than 40 ms (20 ms on
+  the Guition), a contact that moved, a contact that already counted, and a tap within 150 ms of the last accepted
+  tap on the same button (bounce). `show_page` already drops a fill still under way when the next page is asked for,
+  so nothing else changed. `tests/test_cyd_ui.cpp` covers the page buttons.
+- The settings page's own pager (`settings_screen.h`) never used the guard and is unchanged. Tile taps, cards, keys,
+  chips and sliders keep `accept()` and its window.
+
 ### Compatibility 0.2.71 / firmware 0.2.60
 
 Firmware only (`runtime_model.h`, `runtime_tiles.h`); the app raises `FIRMWARE_VERSION`. No protocol, storage or

docs/TEST_RESULTS_0272.md

@@ -0,0 +1,76 @@
+# Test results app 0.2.72 / firmware 0.2.61 (2026-09-18)
+
+Max: "als ik snel naar pagina 4 wil, kan ik niet snel doorklikken, dan moet je steeds wachten totdat de pagina geladen
+is". Next, Next, Next on the page bar took three waits of about half a second each. This release lets the page buttons
+keep up with the finger. See CHANGELOG 0.2.72 and the compatibility note in docs/RELEASING.md. Built on 0.2.71
+(`2cb3479`).
+
+## What happened before
+
+- The `< Previous` and `Next >` bars ran their tap through `cyd::touch_guard.accept()`, the guard of every tile: it
+  drops a second tap on the same tile within 600 ms as a bounce, so one contact can never switch a light twice. On
+  the page bar that window dropped the second and third tap of Next, Next, Next. The page itself was never the
+  problem: since 0.2.42 a page switch places the new page in the next frame and fills its cards two per refresh, and
+  a new switch drops a fill still under way.
+- The -/+ keys of the setpoint pill had the same complaint in 0.2.25 and got `accept_repeat()`: every clean tap
+  counts, only a contact within 150 ms of the last one on the same key is a bounce. The page buttons now use it.
+
+## Automated
+
+- C++ (`clang++ -std=c++17 -Wall -Wextra -Werror -I.`): 17/17 PASS in a clean export of the committed tree plus this
+  round's changes. `tests/test_cyd_ui.cpp` now holds the page buttons: three Next taps 250 ms apart all count, the
+  old `accept()` still refuses within its window, and Previous right after Next counts.
+- Python (`.venv-portal`): 386 tests OK, the same set as 0.2.71 (firmware only; the app raises `FIRMWARE_VERSION`).
+- `generate_packages.py --check` green.
+- The working tree also carried another session's uncommitted `runtime_model.h` / `runtime_tiles.h` work (the
+  48-tile round, 0.2.73); every build and test below ran from a clean export with only this round's changes.
+
+## Firmware on the Mac (host builds, driven over the ESPHome API)
+
+`pager_host.py` (scratchpad, after `wake_sleep_host.py`): the living room layout of the demo home on 8099 (17 tiles,
+four pages), a `pager_taps` action that taps the page bar from an LVGL timer (`count` taps `gap_ms` apart, each a
+90 ms contact ending in `LV_EVENT_SHORT_CLICKED` on the button, as the touch layer produces it), an `old_rule` mode
+that taps the way firmware 0.2.60 did, and a probe that logs the page, the page number label and the fill state.
+**Guition 12/12, CYD 12/12** (the first run judged the bounce and old-rule bursts by the spacing asked for, and
+the host's LVGL loop stalls about a second on the first frame of a burst, macOS throttling the SDL window, so
+two taps meant to be 80 ms apart landed 1040 ms apart; the checks now judge every tap against the spacing that
+really happened, and the second run passed on both boards).
+
+| Check | Guition | CYD |
+|---|---|---|
+| The living room layout shows four pages, page 1 drawn | pass | pass |
+| Next, Next, Next 200 ms apart lands on page 4 (4 / 4) | pass | pass |
+| Every one of those taps moved a page: 2, 3, 4, none refused | pass | pass |
+| Page 4 finishes drawing afterwards | pass | pass |
+| Two more Next taps on the last page change nothing | pass | pass |
+| Previous, Previous, Previous 200 ms apart lands on page 1 | pass | pass |
+| Three Next taps 160 ms apart (six a second) land on page 4 | pass | pass |
+| Three Previous taps 160 ms apart land on page 1 | pass | pass |
+| Bounce: a Next within 150 ms of the last accepted tap is refused, a slower one counts and moves a page | pass | pass |
+| The old rule (0.2.60): a Next within 600 ms of the last accepted tap was refused | pass | pass |
+| Back on page 1 afterwards | pass | pass |
+| No errors in the log | pass | pass |
+
+The bounce burst on the Guition landed at 98, 94 and 81 ms and was accepted, refused, accepted, refused, as the
+rule says (the third tap came 192 ms after the first accepted one); the CYD's at 84, 84 and 96 ms the same way. The
+old-rule burst on the Guition landed at 208, 202 and 203 ms and was accepted, refused, refused, accepted: that is
+what Max felt, one page per 600 ms, however fast he tapped.
+
+## Builds (ESPHome 2026.6.2)
+
+Check profiles with the generated packages and local components; no errors or compiler warnings.
+
+| Profile | RAM | Flash | vs 0.2.71 |
+|---|---|---|---|
+| Guition | 25.7% (84,372 B) | 21.0% (1,708,943 B) | same RAM, -24 B flash |
+| CYD | 24.0% (78,796 B) | 79.2% (1,453,647 B) | same RAM, -36 B flash |
+
+The check builds carry a device name of their own (`check-<board>-f9`) and placeholder Wi-Fi, API and OTA values, so
+the sizes differ from the bench profiles by the length of those strings; the code is the same.
+
+## Not tested
+
+- Real hardware: neither bench screen ran this firmware before the release. Max's own screens get it through Update;
+  the tap that started this is the one to try there, on the Guition and on the CYD.
+- A resistive panel's lift-off bounce on the CYD is covered by the 40 ms minimum contact and the 150 ms gap, the same
+  net the -/+ keys have had since 0.2.25 (physically tested then), not by a new measurement.

guition-4848s040.yaml

@@ -2,7 +2,7 @@
 # Source profile for packages/guition.yaml (tools/generate_packages.py); see docs/GUITION.md.
 # A screen gets its tiles from the ESP Screen Manager add-on while it runs. The CYD is its own profile.
 substitutions:
-  SCREEN_FIRMWARE_VERSION: "0.2.60"
+  SCREEN_FIRMWARE_VERSION: "0.2.61"
   # 0: no drift limit; LVGL decides, releasing inside the tile is a tap, leaving it is not
   # (like iOS). The CYD keeps a limit because its resistive panel jumps.
   TOUCH_MOVE_LIMIT_PX: "0"
@@ -2438,7 +2438,7 @@ lvgl:
             on_short_click:
               - if:
                   condition:
-                    lambda: 'return cyd::touch_guard.accept(millis(), 11);'
+                    lambda: 'return cyd::touch_guard.accept_repeat(millis(), 11);'
                   then:
                     - lambda: 'id(tile_page) = std::max(0, id(tile_page) - 1);'
                     - script.execute: show_tile_page
@@ -2479,7 +2479,7 @@ lvgl:
             on_short_click:
               - if:
                   condition:
-                    lambda: 'return cyd::touch_guard.accept(millis(), 12);'
+                    lambda: 'return cyd::touch_guard.accept_repeat(millis(), 12);'
                   then:
                     - lambda: 'id(tile_page) += 1;'
                     - script.execute: show_tile_page

home-like-2432s028.yaml

@@ -17,7 +17,7 @@
 ############################################################
 
 substitutions:
-  SCREEN_FIRMWARE_VERSION: "0.2.60"
+  SCREEN_FIRMWARE_VERSION: "0.2.61"
   # A tap survives a finger that drifts up to this far (320 px over 58 mm: 1 cm); further is a swipe.
   TOUCH_MOVE_LIMIT_PX: "56"
   # Shortest contact that counts as a tap; a resistive panel bounces, a capacitive one does not.
@@ -2351,7 +2351,7 @@ lvgl:
             on_short_click:
               - if:
                   condition:
-                    lambda: 'return cyd::touch_guard.accept(millis(), 11);'
+                    lambda: 'return cyd::touch_guard.accept_repeat(millis(), 11);'
                   then:
                     - lambda: 'id(tile_page) = std::max(0, id(tile_page) - 1);'
                     - script.execute: show_tile_page
@@ -2392,7 +2392,7 @@ lvgl:
             on_short_click:
               - if:
                   condition:
-                    lambda: 'return cyd::touch_guard.accept(millis(), 12);'
+                    lambda: 'return cyd::touch_guard.accept_repeat(millis(), 12);'
                   then:
                     - lambda: 'id(tile_page) += 1;'
                     - script.execute: show_tile_page

packages/cyd.yaml

@@ -18,7 +18,7 @@
 ############################################################
 
 substitutions:
-  SCREEN_FIRMWARE_VERSION: "0.2.60"
+  SCREEN_FIRMWARE_VERSION: "0.2.61"
   # A tap survives a finger that drifts up to this far (320 px over 58 mm: 1 cm); further is a swipe.
   TOUCH_MOVE_LIMIT_PX: "56"
   # Shortest contact that counts as a tap; a resistive panel bounces, a capacitive one does not.
@@ -2340,7 +2340,7 @@ lvgl:
             on_short_click:
               - if:
                   condition:
-                    lambda: 'return cyd::touch_guard.accept(millis(), 11);'
+                    lambda: 'return cyd::touch_guard.accept_repeat(millis(), 11);'
                   then:
                     - lambda: 'id(tile_page) = std::max(0, id(tile_page) - 1);'
                     - script.execute: show_tile_page
@@ -2381,7 +2381,7 @@ lvgl:
             on_short_click:
               - if:
                   condition:
-                    lambda: 'return cyd::touch_guard.accept(millis(), 12);'
+                    lambda: 'return cyd::touch_guard.accept_repeat(millis(), 12);'
                   then:
                     - lambda: 'id(tile_page) += 1;'
                     - script.execute: show_tile_page

packages/guition.yaml

@@ -3,7 +3,7 @@
 # Source profile for packages/guition.yaml (tools/generate_packages.py); see docs/GUITION.md.
 # A screen gets its tiles from the ESP Screen Manager add-on while it runs. The CYD is its own profile.
 substitutions:
-  SCREEN_FIRMWARE_VERSION: "0.2.60"
+  SCREEN_FIRMWARE_VERSION: "0.2.61"
   # 0: no drift limit; LVGL decides, releasing inside the tile is a tap, leaving it is not
   # (like iOS). The CYD keeps a limit because its resistive panel jumps.
   TOUCH_MOVE_LIMIT_PX: "0"
@@ -2427,7 +2427,7 @@ lvgl:
             on_short_click:
               - if:
                   condition:
-                    lambda: 'return cyd::touch_guard.accept(millis(), 11);'
+                    lambda: 'return cyd::touch_guard.accept_repeat(millis(), 11);'
                   then:
                     - lambda: 'id(tile_page) = std::max(0, id(tile_page) - 1);'
                     - script.execute: show_tile_page
@@ -2468,7 +2468,7 @@ lvgl:
             on_short_click:
               - if:
                   condition:
-                    lambda: 'return cyd::touch_guard.accept(millis(), 12);'
+                    lambda: 'return cyd::touch_guard.accept_repeat(millis(), 12);'
                   then:
                     - lambda: 'id(tile_page) += 1;'
                     - script.execute: show_tile_page

screen_manager/CHANGELOG.md

@@ -1,3 +1,11 @@
+## 0.2.72 (firmware 0.2.61)
+
+Next, Next, Next: the page buttons keep up with your finger.
+
+- **Tap through the pages as fast as you like.** Tap Next three times in a row and you are on page 4, whether or not the pages in between had finished drawing. Before, a second tap on the same button within 600 ms was dropped as a bounce, the rule that keeps a light from switching twice on one tap, so every page made you wait before the next tap counted. The page buttons now follow the rule of the -/+ keys: every clean tap counts, and only a contact within 150 ms of the last one on the same button is taken for a bounce.
+- **Previous does the same**, and tapping past the first or last page still changes nothing.
+- Needs firmware 0.2.61: press **Update** on the screen. Includes everything from 0.2.71.
+
 ## 0.2.71 (firmware 0.2.60)
 
 A small slider stays where you leave it while the lights fade towards it.

screen_manager/app/core.py

@@ -20,7 +20,7 @@ WEEKDAYS = ['Mo', 'Tu', 'We', 'Th', 'Fr', 'Sa', 'Su']
 REPO = 'https://github.com/MaxGramser/homeassistant_espscreen'
 REFS = {'cyd': 'main', 'guition': 'main'}
 # Firmware shipped with this app release; screens below it get an update offer.
-FIRMWARE_VERSION = '0.2.60'
+FIRMWARE_VERSION = '0.2.61'
 # The Auto standby switch a screen offers Home Assistant automations.
 AUTO_STANDBY_MIN_FIRMWARE = '0.2.41'
 # The settings page the screen opens itself, and the screen.settings tile that opens it.

screen_manager/config.yaml

@@ -1,5 +1,5 @@
 name: ESP Screen Manager
-version: "0.2.71"
+version: "0.2.72"
 slug: esp_screen_manager
 description: Choose tiles for your CYD and Guition touchscreen, without reflashing.
 url: https://github.com/MaxGramser/homeassistant_espscreen

tests/test_cyd_ui.cpp

@@ -21,6 +21,14 @@ int main() {
   r.begin(1300); assert(!r.accept_repeat(1340, 7));                                     // 80 ms after the last: bounce
   r.begin(1300); r.update(40, 0); assert(!r.accept_repeat(1500, 7));                   // a swipe is never a step
   r.begin(1600); assert(r.accept_repeat(1660, 8));                                      // the other key right away
+  // Page buttons (0.2.72): Next, Next, Next at a finger's pace all count, so page 4 is three taps away
+  // while the pages are still drawing; accept() would have dropped the second and third (600 ms).
+  cyd::TouchGuard pager;
+  pager.begin(2000); assert(pager.accept_repeat(2080, 12));
+  pager.begin(2250); assert(pager.accept_repeat(2330, 12));                              // 250 ms after the last
+  pager.begin(2500); assert(pager.accept_repeat(2580, 12));                              // and again
+  pager.begin(2600); assert(!pager.accept(2680, 12));                                    // the old rule: bounce window
+  pager.begin(2700); assert(pager.accept_repeat(2780, 11));                              // Previous right after Next
   cyd::TouchGuard g;
   g.begin(100);
   assert(!g.accept(120, 1)); // resistive noise pulse