Collapse mode and layer onto one picker row; fix renderer diagnosis

Only one layer is ever active, so neither mode nor layer needs a scrolling
row. Two pickers side by side cost one 44px row instead of two.

The panel needed that space: .pads uses minmax(34px,1fr) rows and sat
exactly at the 34px floor, so the SP-404 illustration clipped at the
bottom. Measured on device:

  strip   96px -> 44px
  panel  433px -> 485px
  pads   142px -> 194px, rows 34px -> 44.75px
  unit overflow 1px -> 0

The layer list is read out of the reference's own chips at render time,
so its set, ordering and counts cannot drift from renderExplore().

Also corrects the renderer-denial root cause. It is not screen state: a
denial was observed with Reason[screen_on & caller care]. Every failure
followed 'adb install -r' and an immediate relaunch has always succeeded,
so device.sh now retries the force-stop/start cycle instead of blaming
the lockscreen.

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

CLAUDE.md

@@ -41,11 +41,15 @@ ANDROID_HOME=$HOME/Android/Sdk
 
 ## Driving the device
 
-**The screen must be awake and unlocked before launching, every time.** While
+**Launch through `scripts/device.sh`, never a bare `am start`.** TCL denies the
-the display dozes, TCL denies the WebView renderer and the app comes up blank
+WebView renderer in the window just after `adb install -r`, and the app then
-with no crash, no JS error and an empty `document.title`. The failure is sticky
+comes up blank with no crash, no JS error and an empty `document.title`. The
-for the life of the process, so waking the phone afterwards does not repair it —
+failure is sticky for the life of the process, so the app must be force-stopped
-the app has to be force-stopped and started again.
+and started again — which is exactly what the helper retries.
+
+Keep the phone unlocked while working. That is not what causes the denial (see
+Device notes), but nothing on screen can be verified through a lockscreen and
+`input keyevent` will not reach the app.
 
 `scripts/device.sh` handles this so it cannot be forgotten:
 
@@ -142,9 +146,21 @@ they will not come along with the reference:
    them is `display:none` on phones and still in the DOM, so the adjacent-sibling
    selector never matches. Measured gap: 0px. Both selected chips are also
    light-filled and left-aligned, so they visually merge into one shape.
-2. **The 7 context chips collapse into one 44px control** that opens the existing
+2. **Both chip rows collapse into two pickers side by side on one 44px row.**
-   sheet, freeing a row for the panel. The sheet rows show `CTXHINT`, which the
+   Only one layer is ever active, so neither the mode nor the layer needs a
-   phone layout otherwise never displays.
+   scrolling row of its own. Each opens the existing sheet. The mode sheet shows
+   `CTXHINT`, which the phone layout otherwise never displays; the layer sheet
+   shows each layer's function count.
+
+   This matters for the panel, not just for tidiness: `.pads` uses
+   `minmax(34px, 1fr)` rows and on this device it sits exactly at that 34px
+   floor, so the SP-404 illustration clips at the bottom if anything takes more
+   room. Two rows cost 96px; one costs 44px, and the 52px goes to the panel.
+
+   The layer list is read out of the reference's own chips at render time rather
+   than recomputed, so the set, ordering and counts cannot drift from what
+   `renderExplore()` decided. In React this becomes shared derived state
+   instead.
 3. **Chip targets go from 34.9px to 44px.** They scroll horizontally, so an
    undersized target turns a mis-hit into a scroll instead of a tap.
 4. **Overflowing rows get a right-edge fade.** Scrollbars are hidden, and both
@@ -179,22 +195,32 @@ sudo udevadm control --reload-rules && sudo udevadm trigger
 It lives in `/etc/udev` so apt upgrades don't overwrite it. The `uaccess` tag
 grants the logged-in session access, so no logout is needed.
 
-**A blank grey window means the screen was off when the app started.** TCL's
+**A blank grey window means the renderer was denied — relaunch.** TCL's
-`AppBootManagerService` refuses to let the app bind the WebView's sandboxed
+`AppBootManagerService` sometimes refuses to let the app bind the WebView's
-renderer process while the display is dozing:
+sandboxed renderer process:
 
 ```
 W/AppBootManager: Type[TYPE_SERVICE];Result[false];
   Target[org.chromium.content.app.SandboxedProcessService0];
-  Caller[dev.ncode.sp404layers];Reason[screen_off & caller care]
+  Caller[dev.ncode.sp404layers];Reason[screen_on & caller care]
 E/cr_ChildProcessConn: Failed to establish the service connection.
 ```
 
 With no renderer the page never parses, so there is no crash, no JS error, an
 empty `document.title`, and DevTools `Runtime.evaluate` hangs. The failure is
 sticky for the life of the process — reopening from Recents keeps the grey
-window. Wake and unlock the phone first, then `am force-stop` and relaunch.
+window. Only `am force-stop` plus a fresh start clears it.
-This is an OEM policy, not an app bug, and it only bites when launching over adb.
+
+**The trigger is the window just after `adb install -r` replaces the package.**
+Every observed failure followed an install; an immediate relaunch with no
+install has always succeeded on the first try. Do not trust the `Reason` string:
+it reports the current screen state, and both `screen_off` and `screen_on` have
+been seen denying, so screen state is not the cause. An earlier version of this
+file claimed it was — that was wrong.
+
+`scripts/device.sh launch` therefore retries the whole force-stop/start cycle up
+to 3 times. This is an OEM policy, not an app bug, and it only bites when
+launching over adb — tapping the icon is fine.
 
 ## Gotchas
 

native/native.css

@@ -42,6 +42,13 @@
      single .ctxbtn below, leaving one chip row with real touch targets.
      Phase 2 must carry this - see CLAUDE.md. */
 
+  /* Mode and layer pickers share one row: only one layer is ever active, so
+     neither needs a scrolling row of its own. One 44px row instead of two gives
+     the panel back 52px, which it needs - its pad rows sit at their
+     minmax(34px,1fr) floor, so anything less clips the illustration. */
+  .stripbar{ display: flex; gap: 8px; }
+  .stripbar .ctxbtn{ flex: 1 1 0; min-width: 0; font-size: 1rem; padding: 8px 12px; }
+
   .ctxbtn{
     display: flex; align-items: center; justify-content: space-between; gap: 10px;
     width: 100%; min-height: 44px; padding: 8px 14px;

native/native.js

@@ -128,6 +128,47 @@
       .join("");
   }
 
+  /* Snapshot of the reference's own layer chips, refreshed on every render.
+     Read from its DOM rather than recomputed, so the set, ordering and counts
+     cannot drift from what renderExplore() decided. */
+  var layerOptions = [];
+
+  function snapshotLayers(row) {
+    return [].map.call(row.querySelectorAll(".chip"), function (c) {
+      var clone = c.cloneNode(true);
+      var small = clone.querySelector("small");
+      var count = small ? small.textContent.trim() : "";
+      if (small) small.remove();
+      return { layer: c.getAttribute("data-layer") || "", label: clone.textContent.trim(), count: count };
+    });
+  }
+
+  function layerSheetHTML() {
+    var cur = layerKey(st.held);
+    return layerOptions
+      .map(function (o) {
+        return '<button class="ctxopt" data-layer="' + o.layer + '" aria-pressed="' + (o.layer === cur) + '">' +
+          "<b>" + o.label + '</b><span class="hint">' +
+          (o.count === "1" ? "1 function" : o.count + " functions") + " on this layer</span></button>";
+      })
+      .join("");
+  }
+
+  function pickerBtn(sheetAttr, text, aria) {
+    var btn = document.createElement("button");
+    btn.className = "ctxbtn";
+    btn.setAttribute(sheetAttr, "");
+    btn.setAttribute("aria-label", aria);
+    var label = document.createElement("b");
+    label.textContent = text;
+    var caret = document.createElement("span");
+    caret.setAttribute("aria-hidden", "true");
+    caret.textContent = "▾";
+    btn.appendChild(label);
+    btn.appendChild(caret);
+    return btn;
+  }
+
   function reshapeStrip() {
     var strip = document.getElementById("strip");
     if (!strip) return;
@@ -136,18 +177,19 @@
     // Drill renders a .stripbtn and Find an input, so both are left alone.
     var rows = strip.querySelectorAll(".chips");
     if (phone() && rows.length >= 2) {
-      var btn = document.createElement("button");
+      layerOptions = snapshotLayers(rows[1]);
-      btn.className = "ctxbtn";
+
-      btn.setAttribute("data-ctxsheet", "");
+      // Only one layer is ever active, so both rows collapse to pickers. Side by
-      btn.setAttribute("aria-label", "Mode: " + CTX[st.ctx] + ". Change mode");
+      // side they cost one 44px row instead of two, which the panel needs: its
-      var label = document.createElement("b");
+      // pad rows sit at their minmax(34px,1fr) floor and clip below that.
-      label.textContent = CTX[st.ctx];
+      var held = st.held.length ? st.held.map(mName).join(" + ") : "No modifier";
-      var caret = document.createElement("span");
+      var bar = document.createElement("div");
-      caret.setAttribute("aria-hidden", "true");
+      bar.className = "stripbar";
-      caret.textContent = "▾";
+      bar.appendChild(pickerBtn("data-ctxsheet", CTX[st.ctx], "Mode: " + CTX[st.ctx] + ". Change mode"));
-      btn.appendChild(label);
+      bar.appendChild(pickerBtn("data-layersheet", held, "Layer: " + held + ". Change layer"));
-      btn.appendChild(caret);
+
-      rows[0].replaceWith(btn);
+      rows[0].replaceWith(bar);
+      rows[1].remove();
     }
 
     strip.querySelectorAll(".chips").forEach(function (r) {
@@ -171,9 +213,15 @@
       openSheet("Where you are", ctxSheetHTML());
       return;
     }
+    if (t.hasAttribute("data-layersheet")) {
+      openSheet("Layers in " + CTX[st.ctx], layerSheetHTML());
+      return;
+    }
     // The reference's own handler runs first (registered earlier) and applies
-    // data-ctx, so this only has to dismiss the sheet afterwards.
+    // data-ctx / data-layer, so this only has to dismiss the sheet afterwards.
-    if (t.hasAttribute("data-ctx") && t.closest("#sheet")) closeSheet();
+    // data-layer="" (no modifier) is intentional: the reference tests
+    // `d.layer !== undefined`, so an empty value still selects that layer.
+    if ((t.hasAttribute("data-ctx") || t.hasAttribute("data-layer")) && t.closest("#sheet")) closeSheet();
   });
 
   // The first render already ran before this script loaded.

scripts/device.sh

@@ -1,25 +1,32 @@
 #!/usr/bin/env bash
 # Device helpers for driving the app over adb.
 #
-# TCL's AppBootManagerService refuses to let the app bind the WebView's renderer
+# TCL's AppBootManagerService sometimes refuses to let the app bind the
-# process while the display is dozing:
+# WebView's renderer process:
 #
 #   W/AppBootManager: Result[false]; Target[...SandboxedProcessService0];
-#                     Caller[dev.ncode.sp404layers]; Reason[screen_off & caller care]
+#                     Caller[dev.ncode.sp404layers]; Reason[screen_on & caller care]
+#   E/cr_ChildProcessConn: Failed to establish the service connection.
 #
 # With no renderer the page never parses: no crash, no JS error, an empty
 # document.title, and DevTools Runtime.evaluate hangs. It is sticky for the life
 # of the process, so reopening from Recents keeps the blank window.
 #
-# So: always ensure the screen is awake before launching, and say so plainly
+# Observed trigger is the window just after `adb install -r` replaces the
-# when it is not, rather than leaving a blank window to puzzle over.
+# package - every failure followed an install, and an immediate relaunch with no
+# install has always succeeded. The Reason string reports the current screen
+# state ("screen_on" and "screen_off" have both been seen denying), so screen
+# state is not the cause. Hence: launch retries the whole cycle.
 set -euo pipefail
 
 ADB="${ADB:-$HOME/Android/Sdk/platform-tools/adb}"
 PKG=dev.ncode.sp404layers
 ACT="$PKG/.MainActivity"
 PORT=9333
+ATTEMPTS=3
 
+# Not the cause of the denial, but a locked phone cannot be checked visually
+# and input keyevents will not reach the app.
 ensure_awake() {
   "$ADB" shell input keyevent KEYCODE_WAKEUP >/dev/null 2>&1 || true
   "$ADB" shell input keyevent 82 >/dev/null 2>&1 || true
@@ -30,14 +37,11 @@ ensure_awake() {
 
   if [ "$wake" != "Awake" ] || [ "$lock" = "true" ]; then
     echo "" >&2
-    echo "  ####################################################################" >&2
+    echo "  #############################################################" >&2
     echo "  #  UNLOCK THE PHONE                                         #" >&2
-    echo "  #                                                                  #" >&2
     echo "  #  wakefulness=$wake  lockscreen=$lock" >&2
-    echo "  #                                                                  #" >&2
+    echo "  #  Nothing on screen can be verified while it is locked.    #" >&2
-    echo "  #  While the screen is off, TCL denies the WebView renderer and    #" >&2
+    echo "  #############################################################" >&2
-    echo "  #  the app shows a blank window with no error anywhere.            #" >&2
-    echo "  ####################################################################" >&2
     echo "" >&2
     return 1
   fi
@@ -54,18 +58,16 @@ sleep_normal() {
   echo "stay-awake: reverted to system default"
 }
 
-# Force-stop, start, and wait until the page is actually parsed - not merely
+# Wait until the page is actually parsed, not merely until the activity is up:
-# until the activity is up, which happens even when the renderer was denied.
+# the activity starts fine even when the renderer was denied.
-launch() {
+page_loaded() {
-  ensure_awake || return 1
-  "$ADB" shell am force-stop "$PKG"
-  "$ADB" shell am start -W -n "$ACT" >/dev/null
-
   local pid sock i
-  # Read the pid AFTER starting: an `adb install -r` kills the old process, and
+  # Read the pid AFTER starting. `adb install -r` kills the old process, so a pid
-  # a stale pid makes the devtools socket lookup fail while the app is fine.
+  # captured earlier points at a dead one and the socket lookup fails while the
+  # app itself is healthy - that produced one false "page never loaded".
   pid=$("$ADB" shell pidof "$PKG" | tr -d '\r')
-  for i in $(seq 1 60); do
+  [ -n "$pid" ] || return 1
+  for i in $(seq 1 40); do
     sock=$("$ADB" shell cat /proc/net/unix | grep -o "webview_devtools_remote_$pid" | head -1)
     if [ -n "$sock" ]; then
       "$ADB" forward "tcp:$PORT" "localabstract:$sock" >/dev/null 2>&1 || true
@@ -75,9 +77,23 @@ launch() {
       fi
     fi
   done
+  return 1
+}
 
-  echo "ERROR: activity started but the page never loaded." >&2
+launch() {
-  echo "Renderer denials in the log:" >&2
+  ensure_awake || return 1
+  local attempt
+  for attempt in $(seq 1 $ATTEMPTS); do
+    "$ADB" shell am force-stop "$PKG"
+    "$ADB" shell am start -W -n "$ACT" >/dev/null
+    if page_loaded; then
+      [ "$attempt" -gt 1 ] && echo "(took $attempt attempts - post-install renderer denial)"
+      return 0
+    fi
+    echo "attempt $attempt: renderer denied, retrying" >&2
+  done
+
+  echo "ERROR: page never loaded after $ATTEMPTS attempts." >&2
   "$ADB" logcat -d | grep -E 'AppBootManager.*Result\[false\]|ChildProcessConn: Failed' | tail -4 >&2
   return 1
 }