Fix NModeStepCmd's ret=false bug; attempt (unconfirmed) fix for animation radio box not switching (#81)

NModeStepCmd::execute() declared `ret = false` and never set it true on
success, unlike ShowFragCmd/NModeTraceCmd -- fixed to match. Found while
investigating a report that vibration-mode animation (#81, supposedly
fixed by 07e7bf9) still didn't work live; doesn't appear to be the
actual cause (Builder::execute() calls refreshRenderArea() unconditionally
regardless of a command's return value) but is a real, distinct defect.

Live debugging (strace, live screenshots via xdotool, a synced synthetic
click) traced the real symptom to something earlier: the Vibrational
Frequencies panel's Animation/Vector radio box doesn't deliver its click
event under wx3.2/GTK3 here. Confirmed at the syscall level with an
instrumented build -- the native GTK widget's selected bullet toggles
correctly on click, but NModePanel::OnRadioboxSelected() is never
entered, for a real click or a precisely-synced synthetic one. Since the
Animation/Vector display switch (row visibility, scene-graph switches)
is entirely gated on that event, nothing downstream ever runs.

Added NModePanel::OnRadioboxUpdateUI(), bound via EVT_UPDATE_UI, on the
theory that wx's idle-time UI-update poll would sidestep whatever GTK
signal wiring is failing for the click event specifically. Built,
packaged, and live-tested same day: no visible difference. This fix
does NOT resolve the reported symptom, or doesn't for the reason
assumed. Left in place (harmless, and correct in isolation) with an
honest code comment and a CLAUDE.md writeup for the next session, since
the diagnostic work (confirming the native event never arrives) is
solid and shouldn't need re-deriving, but the fix itself is not.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4kYrC9KXyyiUzQrkS8ha3

Author: andy

CLAUDE.md

@@ -349,6 +349,44 @@ Both per-user, both non-root, both started automatically by the
   that a symptom is this bug rather than a data problem. Fixed with a
   pending-redraw flag (`p_redrawPending`) that `OnPaint()` checks and
   acts on once the in-flight paint finishes.
+- **OPEN, UNRESOLVED (2026-09-17): the Vibrational Frequencies panel's
+  Animation/Vector radio box doesn't deliver its click event under
+  wx3.2/GTK3** — confirmed live via strace (syscall-level, a synthetic
+  click synced exactly with the trace window): the native GTK widget's
+  own selected bullet toggles correctly, but `NModePanel::
+  OnRadioboxSelected()` is never entered — zero evidence of it running,
+  for a real click or a synced synthetic one. This is *not* the same
+  bug as the `p_isValid`/`CreateGrid()` construction-ordering issue
+  above (which is fixed) — this is later, after the panel is fully
+  built and idle, on an ordinary click. Symptom: switching the radio
+  does nothing (no row swap, no vector-arrows/animate-mode switch,
+  Play button never appears) since the whole display-mode switch is
+  gated on that event firing. An `EVT_UPDATE_UI`-based idle-poll
+  workaround was added (`NModePanel::OnRadioboxUpdateUI`, compares
+  `radbox->GetSelection()` each idle tick against a cached
+  `p_lastRadioSel`) on the theory that this sidesteps whatever GTK
+  signal wiring is failing for the click event specifically. Built,
+  packaged, and live-tested same day — **no visible difference**, so
+  this fix does not actually work, or doesn't work for the reason
+  assumed. Don't trust the code comment above `OnRadioboxUpdateUI()`
+  to mean this is fixed — re-verify live before believing it. Next
+  session should treat the "native event never fires" diagnosis as
+  solid (already re-derived twice, syscall-level) but budget fresh
+  investigation for *why* the workaround didn't help — candidates not
+  yet ruled out: whether `EVT_UPDATE_UI` is actually reaching this
+  window at all (same category of failure as the click event, would
+  need its own strace/instrumented-build check), whether
+  `showAnimationMode()`/`showVectorMode()` are even the code path
+  actually driving what's visible in the panel (unverified assumption
+  going in), or a packaging/install mismatch (binary in the `.deb`
+  not matching what was last built — check md5sum/mtime of
+  `/opt/ecce/bin/builder` against `build-cmake/builder` before
+  re-testing anything, this bit a previous round in the same session).
+  This also blocks separately verifying whether `NModeStepCmd`'s
+  redraw fix (`07e7bf9` plus this session's `ret=true` correction)
+  actually animates once display-mode switching works at all — that
+  is *still unverified end-to-end* despite the code-level trace in
+  `NModeStepCmd.C`/`Builder::execute()` looking sound.
 
 ## Where the history lives
 This file used to be a session-by-session diary and grew to ~2500 lines.

src/apps/builder/NModePanel.C

@@ -62,6 +62,7 @@ BEGIN_EVENT_TABLE( NModePanel, NModesGUI )
     EVT_GRID_SELECT_CELL(NModePanel::OnModeSelection)
 
     EVT_RADIOBOX( ID_RADIOBOX_NMODE_VIZTYPE, NModePanel::OnRadioboxSelected )
+    EVT_UPDATE_UI( ID_RADIOBOX_NMODE_VIZTYPE, NModePanel::OnRadioboxUpdateUI )
     EVT_TIMER(wxID_ANY, NModePanel::OnTimer)
     EVT_PLOTCTRL_CLICKED(wxID_ANY, NModePanel::OnPlotClick)
     EVT_PLOTCTRL_POINT_CLICKED(wxID_ANY, NModePanel::OnPointClick)
@@ -84,7 +85,8 @@ NModePanel::NModePanel()
     p_currentStep(0),
     p_mode(0),
     p_numAnimations(0),
-    p_isValid(false)
+    p_isValid(false),
+    p_lastRadioSel(-1)
 {
    p_vecAmplitude = 1.0;
    p_aniAmplitude = 1.0;
@@ -105,7 +107,8 @@ NModePanel::NModePanel(IPropCalculation *calculation,
     p_currentStep(0),
     p_mode(0),
     p_numAnimations(0),
-    p_isValid(false)
+    p_isValid(false),
+    p_lastRadioSel(-1)
 {
    Create(calculation, parent, id, pos, size, style, name);
    p_vecAmplitude = 1.0;
@@ -701,7 +704,8 @@ void NModePanel::OnSliderTextEnter(wxCommandEvent& event)
 void NModePanel::receiveFocus()
 {
    wxRadioBox *radbox = (wxRadioBox*)FindWindow(ID_RADIOBOX_NMODE_VIZTYPE);
-   if (radbox->GetSelection() == 0) {
+   p_lastRadioSel = radbox->GetSelection();
+   if (p_lastRadioSel == 0) {
       showAnimationMode();
    } else {
       showVectorMode();
@@ -779,15 +783,56 @@ void NModePanel::OnCheckboxNmodeVecsignClick( wxCommandEvent& event )
 void NModePanel::OnRadioboxSelected( wxCommandEvent& event )
 {
    wxRadioBox *radbox = (wxRadioBox*)FindWindow(ID_RADIOBOX_NMODE_VIZTYPE);
+   p_lastRadioSel = radbox->GetSelection();
    setSlider();
-   if (radbox->GetSelection() == 0) {
+   if (p_lastRadioSel == 0) {
       showAnimationMode();
    } else {
       showVectorMode();
    }
    event.Skip();
 }
 
+/**
+ * Confirmed live (strace on an instrumented build, synthetic click
+ * synced with the trace window): clicking the Animation/Vector radio
+ * box toggles the native GTK widget's own selected bullet, but
+ * OnRadioboxSelected() below is never entered -- zero writes from any
+ * of its instrumented call sites, for either the real click or a
+ * synced synthetic one. wxEVT_COMMAND_RADIOBOX_SELECTED is not
+ * reaching this object under wx3.2/GTK3 for this control.
+ *
+ * This handler was added as a suspected fix: EVT_UPDATE_UI is a
+ * wx-level idle-time poll, independent of whatever GTK signal wiring
+ * is failing for the click event, so in theory it should still detect
+ * radbox->GetSelection() changing even when the native event doesn't
+ * arrive. Built, packaged, and live-tested (2026-09-17) -- no visible
+ * difference; the Animation/Vector display still doesn't switch. That
+ * means either this EVT_UPDATE_UI handler also isn't firing for this
+ * control, or the display-switch code path itself (showAnimationMode/
+ * showVectorMode) isn't the actual mechanism behind what's visible in
+ * the panel, or the installed package still wasn't picking up this
+ * binary for some reason not yet ruled out. Not root-caused. Next
+ * session: don't assume this fix works, and don't re-spend time
+ * re-deriving the "native event never arrives" fact above -- it's
+ * solid (syscall-level trace, synced click) -- but treat the
+ * EVT_UPDATE_UI theory and everything below it as unverified.
+ */
+void NModePanel::OnRadioboxUpdateUI( wxUpdateUIEvent& event )
+{
+   wxRadioBox *radbox = (wxRadioBox*)FindWindow(ID_RADIOBOX_NMODE_VIZTYPE);
+   int sel = radbox->GetSelection();
+   if (sel != p_lastRadioSel) {
+      p_lastRadioSel = sel;
+      setSlider();
+      if (sel == 0) {
+         showAnimationMode();
+      } else {
+         showVectorMode();
+      }
+   }
+}
+
 void NModePanel::OnModeSelection( wxGridEvent& event )
 {
    // wxGrid::CreateGrid() (NModesGUI::CreateControls(), building the

src/apps/builder/NModePanel.H

@@ -79,6 +79,7 @@ class NModePanel : public NModesGUI,
       void OnEndSliderMotion(wxScrollEvent& event);
       void OnSliderTextEnter(wxCommandEvent& event);
       void OnRadioboxSelected( wxCommandEvent& event );
+      void OnRadioboxUpdateUI( wxUpdateUIEvent& event );
       void OnModeSelection( wxGridEvent& event );
       void OnCheckboxNmodeVecsignClick( wxCommandEvent& event );
       void OnTimer( wxTimerEvent& event );
@@ -104,6 +105,7 @@ class NModePanel : public NModesGUI,
       int         p_mode;
       int         p_numAnimations;
       bool        p_isValid;
+      int         p_lastRadioSel;
       float       p_vecAmplitude; // varies for vector vs animation
       float       p_aniAmplitude; // varies for vector vs animation
 

src/viz/propsgcommands/NModeStepCmd.C

@@ -71,6 +71,8 @@ bool NModeStepCmd::execute()
 
    if (nmtrace) {
 
+      ret = true;
+
       int table = 0 ;  // we only have one table
 
       // reset the coordinates in the fragment for each atom