GPU correctness fixes and optimizations

draw/examples/hellope.odin

@@ -279,12 +279,14 @@ hellope_custom :: proc() {
 	}
 
 	gauge := Gauge {
-		value = 0.73,
-		color = {50, 200, 100, 255},
+		value    = 0.73,
+		color    = {50, 200, 100, 255},
+		bg_color = {80, 80, 80, 255},
 	}
 	gauge2 := Gauge {
-		value = 0.45,
-		color = {200, 100, 50, 255},
+		value    = 0.45,
+		color    = {200, 100, 50, 255},
+		bg_color = {80, 80, 80, 255},
 	}
 
 	// `clay.CustomElementConfig.customData` is a rawptr; the Clay integration in `draw`
@@ -342,11 +344,11 @@ hellope_custom :: proc() {
 			// reflection inside the strip), and gauge2 is deferred-replayed by
 			// `prepare_clay_batch` after the bracket closes (renders crisp on top of the
 			// bracket output — unrelated to the strip since they don't overlap).
+			// `backgroundColor` is omitted on the gauges; bg lives on `Gauge.bg_color`. See `draw_custom`.
 			if clay.UI(clay.ID("gauge"))(
 			{
 				layout = {sizing = {clay.SizingFixed(300), clay.SizingFixed(30)}},
 				custom = {customData = &gauge_custom},
-				backgroundColor = {80, 80, 80, 255},
 			},
 			) {
 				if clay.UI(clay.ID("backdrop"))(
@@ -362,7 +364,6 @@ hellope_custom :: proc() {
 			{
 				layout = {sizing = {clay.SizingFixed(300), clay.SizingFixed(30)}},
 				custom = {customData = &gauge2_custom},
-				backgroundColor = {80, 80, 80, 255},
 			},
 			) {}
 		}
@@ -376,8 +377,9 @@ hellope_custom :: proc() {
 	}
 
 	Gauge :: struct {
-		value: f32,
-		color: draw.Color,
+		value:    f32,
+		color:    draw.Color,
+		bg_color: draw.Color,
 	}
 
 	draw_custom :: proc(layer: ^draw.Layer, bounds: draw.Rectangle, render_data: clay.CustomRenderData) {
@@ -386,14 +388,12 @@ hellope_custom :: proc() {
 		// before the union refactor.
 		gauge := cast(^Gauge)render_data.customData
 
+		// `gauge.bg_color` instead of `render_data.backgroundColor`: under Clay master, an
+		// element with both `custom.customData` and `backgroundColor` emits a Custom AND a
+		// Rectangle for the same bounds, in that order — the Rectangle paints over the
+		// callback's output. Carrying bg on user data sidesteps it.
 		border_width: f32 = 2
-		draw.rectangle(
-			layer,
-			bounds,
-			draw.color_from_clay(render_data.backgroundColor),
-			outline_color = draw.WHITE,
-			outline_width = border_width,
-		)
+		draw.rectangle(layer, bounds, gauge.bg_color, outline_color = draw.WHITE, outline_width = border_width)
 
 		fill := draw.Rectangle {
 			x      = bounds.x,