The three multibar reds were one duplicate load, not a loader defect

`_G["MultiBarBottomLeftButton1NormalTexture"]` really was a different object
from that button's `GetNormalTexture()`, with both answering that same
`GetName()`, so every tint written through the global — the realm's own
`ActionButton_ShowGrid`, and pfUI constantly — landed on a region nothing drew.
I wrote that up as a state-texture publish naming the wrong region at two levels
of template inheritance and left three tests red on it.

It was neither the publish nor the depth. **These tests loaded
`MultiActionBars.xml` twice**: once through `load_action_bar`, where toc order
puts it, and again in each test's own list, left behind when the first was added
— 15 tests. Two loads build two sets of MultiBar buttons, and the two `_G`
publishes disagree about a collision: a frame's is non-overwriting (`0x701bd0`),
so `MultiBarBottomLeftButton1` stayed the *first* set's button, while the
loader's state-texture publish is a plain `globals().set`, so the region name
became the *second* set's texture. Every write through the global reached the
hidden second set.

What rules the loader out is now its own test: `expand`'s merge concatenates
children, and only the base of the real chain declares a `<NormalTexture>` —
`ActionBarButtonTemplate` and `MultiBar1ButtonTemplate` declare only `<Scripts>`,
read off the chain — so there is exactly one declaration at any depth, and a
synthetic three-level chain answers `named == live` at every level, top-level or
child of a named parent.

The other two reds in that file were a misread of the reference, in the same
place twice: `ShapeshiftBar_Update`'s last line is
`UIParent_ManageFramePositions()`, and the pass hides all three shelf strips
under `if ( MultiBarBottomLeft:IsShown() )`. So a form learned while the
bottom-left bar is up shows the middle strip and takes it down again inside one
call — it is never seen across the row, where both tests asserted it would be
"until the next pass".

17/17 in that module now. The publish asymmetry is left alone and written down
where it was found: no single load can reach it, the reference's region-name path
is not disassembled, and matching frames on a guess is not a fix.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SgstFWtn1RxEJ5kVrpgYwn

crates/benilla-app/src/ui_script/multibar_stance_tests.rs

@@ -275,7 +275,6 @@ fn shipped_stance_bar_drives_end_to_end() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     show_bars(&s, &[1]);
     assert!(
         s.eval::<bool>("return ShapeshiftBarFrame ~= nil and ShapeshiftButton10 ~= nil")
@@ -550,7 +549,6 @@ fn the_vertical_multibars_exist_hidden_on_the_reference_pages() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
 
     for bar in ["MultiBarRight", "MultiBarLeft"] {
         assert!(
@@ -605,7 +603,6 @@ fn every_extra_bar_stays_down_until_its_own_toggle_is_set() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
 
     const BARS: [&str; 4] = [
         "MultiBarBottomLeft",
@@ -682,7 +679,6 @@ fn raising_a_bottom_bar_moves_the_managed_bottom_stack() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     load_xml(&s, "Interface\\FrameXML\\CastingBarFrame.xml");
 
     // Read the row's y off the anchor, so a hidden cast bar answers too.
@@ -741,7 +737,6 @@ fn viewable_action_bar_pages_follow_the_bar_toggles() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
 
     let viewable = |s: &UiScript| {
         s.eval::<String>(
@@ -803,7 +798,6 @@ fn the_grid_option_holds_the_extra_bars_empty_wells_open() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     show_bars(&s, &[1]);
 
     let well = |s: &UiScript| {
@@ -884,7 +878,6 @@ fn a_held_payload_ghosts_the_empty_wells_it_opens() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     s.set_action(
         1,
         Some(ActionSlot {
@@ -939,21 +932,26 @@ fn a_held_payload_ghosts_the_empty_wells_it_opens() {
         23,
         "11 empty main wells swap art + MultiBarBottomLeft's 12 appear"
     );
-    // **Known red, and precisely located.** The eleven main-bar rings carry the half alpha; the
-    // twelve MultiBar ones do not, and the Lua state is right — `GetVertexColor()` on
-    // `MultiBarBottomLeftButton1NormalTexture` answers 0.5. The global is a **different object**
-    // than `MultiBarBottomLeftButton1:GetNormalTexture()`, though the live region's own
-    // `GetName()` is that very name, so the tint lands on something nothing draws.
+    // This assertion was red for a while on a wrong diagnosis worth leaving written down, because
+    // the symptom is a good imitation of a loader bug. The twelve MultiBar rings were not taking
+    // the tint, and `_G["MultiBarBottomLeftButton1NormalTexture"]` was a **different object** from
+    // `MultiBarBottomLeftButton1:GetNormalTexture()` while both answered that same `GetName()` —
+    // which reads exactly like the state-texture publish naming the wrong region, and which I
+    // wrote up as a template-depth defect in `loader::widgets`.
     //
-    // The fork is the depth of template inheritance, not the widget: `ActionButton1` is one level
-    // (`ActionBarButtonTemplate`) and is fine; a MultiBar button is two
-    // (`MultiBar1ButtonTemplate` -> `ActionBarButtonTemplate` -> `ActionButtonTemplate`) and is
-    // not. Neither of the two outer templates declares a `<NormalTexture>` of its own, so the
-    // duplicate comes out of the loader's expansion, not the markup.
+    // It was neither. **This test loaded `MultiActionBars.xml` twice** — once through
+    // `load_action_bar`, where it belongs by toc order, and once again in its own list, left
+    // behind when the first was added. Two loads build two sets of MultiBar buttons, and the two
+    // `_G` publishes disagree about a collision: a frame's is non-overwriting (`0x701bd0`), so
+    // `MultiBarBottomLeftButton1` stayed the **first** set's button, while the loader's
+    // state-texture publish is a plain `globals().set`, so the region name became the **second**
+    // set's texture. Every write through the global landed on the hidden second set.
     //
-    // Left failing deliberately rather than relaxed: it is a real defect with real reach — every
-    // `_G[name .. "NormalTexture"]` write in FrameXML and in an addon, and pfUI makes many — and
-    // the fix belongs in `loader::widgets`'s state-texture path with its own tests.
+    // A minimal three-level template chain does *not* reproduce it (`loader/tests.rs`), which is
+    // what finally killed the depth story. The inconsistency between the two publishes is real
+    // but out of reach of any single load, so it is recorded here rather than guessed at: the
+    // reference's region path is not disassembled, and making regions non-overwriting to match
+    // frames would be a change with no evidence behind it.
     assert!(
         ghosts.iter().all(|a| *a == 0.5),
         "every grid ring is the ref's half-alpha ghost, not an opaque plate: {ghosts:?}"
@@ -1022,7 +1020,6 @@ fn a_bar_toggle_sends_the_byte_its_globals_pack_to() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     let _ = s.take_action_bar_toggle_sends();
 
     // The Options row's sequence: assign, update the bars, send.
@@ -1099,7 +1096,6 @@ fn the_shipped_setter_passes_exactly_four_arguments() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
 
     // Lua 5.0's `arg.n`: `...` as a value is not in the 1.12 client's grammar.
     s.run(
@@ -1145,7 +1141,6 @@ fn the_stance_bar_sits_where_the_pass_puts_it() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     load_xml(&s, "Interface\\FrameXML\\Cooldown.xml");
     load_xml(&s, "Interface\\FrameXML\\ActionButtonTemplate.xml");
     s.set_shapeshift_forms(vec![ShapeshiftFormView {
@@ -1206,7 +1201,6 @@ fn the_stance_shelf_follows_the_bottom_left_bar() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     load_xml(&s, "Interface\\FrameXML\\Cooldown.xml");
     load_xml(&s, "Interface\\FrameXML\\ActionButtonTemplate.xml");
 
@@ -1270,15 +1264,21 @@ fn the_stance_shelf_follows_the_bottom_left_bar() {
     }
     assert_eq!(ring(&s), 50.0, "raised rings are the ref's 50");
 
-    // A third form learned while raised: `ShapeshiftBar_Update` shows the middle strip regardless
-    // (BonusActionBarFrame.lua:169-175) and nothing runs the pass on that event, so the strip
-    // draws over the row until the next pass (UIParent.lua:1706-1717).
+    // A third form learned while raised. `ShapeshiftBar_Update`'s form-count arm *does* show the
+    // middle strip — and then its own last line runs the manage pass, which takes it straight back
+    // down because the bottom-left bar is up (`BonusActionBarFrame.lua`'s `ShapeshiftBar_Update`
+    // ends `ShapeshiftBar_UpdateState(); UIParent_ManageFramePositions();`, and the pass's
+    // shapeshift arm hides all three strips under `if ( MultiBarBottomLeft:IsShown() )`). So the
+    // strip never draws across the row, not even for a frame: the show and the hide are inside one
+    // call. This assertion used to read the other way, on a note that "nothing runs the pass on
+    // that event" — the file's last line says otherwise, and that was the whole of this red.
     s.set_shapeshift_forms(vec![form(2457), form(71), form(2458)]);
     s.fire_event("UPDATE_SHAPESHIFT_FORMS", vec![]);
     assert!(
-        shown(&s, "ShapeshiftBarMiddle"),
-        "Update shows the strip; only the pass takes it down"
+        !shown(&s, "ShapeshiftBarMiddle"),
+        "Update shows the strip and its own trailing pass hides it again"
     );
+    // And an explicit pass is idempotent on that state.
     s.run("UIParent_ManageFramePositions()").unwrap();
     assert!(!shown(&s, "ShapeshiftBarMiddle"));
     assert_eq!(ring(&s), 50.0);
@@ -1318,7 +1318,6 @@ fn the_stance_shelf_is_as_long_as_the_form_count() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     load_xml(&s, "Interface\\FrameXML\\Cooldown.xml");
     load_xml(&s, "Interface\\FrameXML\\ActionButtonTemplate.xml");
     show_bars(&s, &[]); // unraised: the shelf art is the state under test
@@ -1418,7 +1417,6 @@ fn an_extra_bars_empty_well_keeps_its_bound_hotkey_label() {
     load_xml(&s, r"Interface\FrameXML\LocaleProperties.lua");
     load_xml(&s, "KeyBindingsPage.xml");
     load_xml(&s, "OptionsFrame.xml");
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     show_bars(&s, &[1]);
     s.set_action(
         61,
@@ -1500,7 +1498,6 @@ fn the_middle_strip_tiles_along_its_length_only() {
     let mut s = UiScript::new().unwrap();
     s.set_screen_size(1024.0, 768.0);
     load_action_bar(&s);
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     show_bars(&s, &[]); // bars down: the shelf is the state under test
 
     let form = |id: u32| ShapeshiftFormView {
@@ -1567,7 +1564,6 @@ fn a_forms_state_change_leaves_the_shelf_down_over_the_raised_bar() {
     let mut s = UiScript::new().unwrap();
     s.set_screen_size(1024.0, 768.0);
     load_action_bar(&s);
-    load_xml(&s, "Interface\\FrameXML\\MultiActionBars.xml");
     // A recorder for the two events a push may announce.
     s.run(
         "STANCE_LOG = {} local f = CreateFrame('Frame') \
@@ -1721,8 +1717,11 @@ fn a_forms_state_change_leaves_the_shelf_down_over_the_raised_bar() {
     );
     assert_eq!(log(&s), "");
 
-    // A fourth stance learned is a list edge: `ShapeshiftBar_Update` shows the strip over the
-    // raised bar until the next pass.
+    // A fourth stance learned is a list edge. `ShapeshiftBar_Update` shows the strip for the form
+    // count and then runs the manage pass on its own last line, which takes it down again under
+    // the raised bar — so the shelf is never seen across the row. Same correction as
+    // `the_stance_shelf_follows_the_bottom_left_bar`'s: the "until the next pass" reading ignored
+    // `ShapeshiftBar_Update`'s trailing `UIParent_ManageFramePositions()`.
     assert_eq!(
         push_forms(
             &mut s,
@@ -1738,8 +1737,8 @@ fn a_forms_state_change_leaves_the_shelf_down_over_the_raised_bar() {
     );
     assert_eq!(log(&s), "UPDATE_SHAPESHIFT_FORMS");
     assert!(
-        shown(&s, "ShapeshiftBarMiddle"),
-        "the learn edge: Update shows the strip"
+        !shown(&s, "ShapeshiftBarMiddle"),
+        "the learn edge shows the strip and hides it again in the same call"
     );
     s.run("UIParent_ManageFramePositions()").unwrap();
     assert!(!shown(&s, "ShapeshiftBarMiddle"));

crates/benilla-ui/src/loader/tests.rs

@@ -14,6 +14,82 @@ mod loader_tests {
         framexml::parse(text).expect("valid FrameXML")
     }
 
+    /// A state texture's `$parent` name points at the live slot however deep the template chain
+    /// is, and whether the button is top-level or a named parent's child.
+    ///
+    /// Written after a wrong diagnosis: `_G["MultiBarBottomLeftButton1NormalTexture"]` really was
+    /// a different object from that button's `GetNormalTexture()`, and the obvious reading was
+    /// that `expand`'s merge duplicates the base's `<NormalTexture>` once per level, so the
+    /// publish and the final slot drift apart. It does not — merge concatenates children and only
+    /// the base of the real chain (`ActionButtonTemplate`) declares one, so there is exactly one
+    /// declaration at any depth. The real cause was the test loading `MultiActionBars.xml` twice
+    /// (`multibar_stance_tests::a_held_payload_ghosts_the_empty_wells_it_opens` carries the
+    /// write-up). This test is what rules the loader out, so it stays.
+    #[test]
+    fn a_state_textures_name_follows_the_live_slot_at_any_template_depth() {
+        let mut s = UiScript::new().unwrap();
+        s.set_screen_size(800.0, 600.0);
+        let doc = parse(
+            r#"<Ui>
+                <CheckButton name="DepthBase" virtual="true">
+                    <Size><AbsDimension x="36" y="36"/></Size>
+                    <NormalTexture name="$parentNormalTexture" file="Interface\Base">
+                        <Size><AbsDimension x="66" y="66"/></Size>
+                        <Anchors><Anchor point="CENTER"/></Anchors>
+                    </NormalTexture>
+                </CheckButton>
+                <!-- Two more levels that declare no state texture of their own, as
+                     ActionBarButtonTemplate and MultiBar1ButtonTemplate do not. -->
+                <CheckButton name="DepthMid" virtual="true" inherits="DepthBase">
+                    <Scripts><OnLoad>this.mid = 1</OnLoad></Scripts>
+                </CheckButton>
+                <CheckButton name="DepthOuter" virtual="true" inherits="DepthMid">
+                    <Scripts><OnLoad>this.outer = 1</OnLoad></Scripts>
+                </CheckButton>
+                <CheckButton name="OneHop" inherits="DepthBase">
+                    <Anchors><Anchor point="TOPLEFT"/></Anchors>
+                </CheckButton>
+                <CheckButton name="ThreeHop" inherits="DepthOuter">
+                    <Anchors><Anchor point="TOPRIGHT"/></Anchors>
+                </CheckButton>
+                <Frame name="DepthHost">
+                    <Anchors><Anchor point="BOTTOMLEFT"/></Anchors>
+                    <Frames>
+                        <CheckButton name="$parentKid" inherits="DepthOuter">
+                            <Anchors><Anchor point="CENTER"/></Anchors>
+                        </CheckButton>
+                    </Frames>
+                </Frame>
+            </Ui>"#,
+        );
+        let report = load(&s, &doc, &no_files);
+        assert!(report.errors.is_empty(), "{:?}", report.errors);
+        for button in ["OneHop", "ThreeHop", "DepthHostKid"] {
+            assert!(
+                s.eval::<bool>(&format!(
+                    "return _G['{button}NormalTexture'] == _G['{button}']:GetNormalTexture()"
+                ))
+                .unwrap(),
+                "{button}: the published name is not the live normal texture"
+            );
+            assert_eq!(
+                s.eval::<String>(&format!(
+                    "return _G['{button}']:GetNormalTexture():GetName()"
+                ))
+                .unwrap(),
+                format!("{button}NormalTexture"),
+                "{button}: the live region does not answer its own published name"
+            );
+            // And the paint reached it, so this is not two agreeing empties.
+            assert_eq!(
+                s.eval::<String>(&format!("return _G['{button}NormalTexture']:GetTexture()"))
+                    .unwrap(),
+                "Interface\\Base",
+                "{button}: the base template's art"
+            );
+        }
+    }
+
     /// `CMinimap::LoadXML 0x4ee2b0`: `minimapArrowModel` goes to engine children 1-8 (`0x4ee170`),
     /// `minimapPlayerModel` to child 9 alone (`[Minimap+0x338]`, `0x4ee260`); without the
     /// attributes the defaults are `0x84c768` and `0x8453c0`. `Minimap.xml:97` sets both.