internal/shader: better error message for the fragment entry point

Closes #3073

image.go

@@ -295,6 +295,16 @@ type Vertex struct {
 	ColorG float32
 	ColorB float32
 	ColorA float32
+
+	// Custom0/Custom1/Custom2/Custom3 represents general-purpose values passed to the shader.
+	// In order to use them, Fragment must have an additional vec4 argument.
+	//
+	// These values are valid only when DrawTrianglesShader is used.
+	// In other cases, these values are ignored.
+	Custom0 float32
+	Custom1 float32
+	Custom2 float32
+	Custom3 float32
 }
 
 var _ [0]byte = [unsafe.Sizeof(Vertex{}) - unsafe.Sizeof(float32(0))*graphics.VertexFloatCount]byte{}
@@ -668,6 +678,10 @@ func (i *Image) DrawTrianglesShader(vertices []Vertex, indices []uint16, shader
 		vs[i*graphics.VertexFloatCount+5] = v.ColorG
 		vs[i*graphics.VertexFloatCount+6] = v.ColorB
 		vs[i*graphics.VertexFloatCount+7] = v.ColorA
+		vs[i*graphics.VertexFloatCount+8] = v.Custom0
+		vs[i*graphics.VertexFloatCount+9] = v.Custom1
+		vs[i*graphics.VertexFloatCount+10] = v.Custom2
+		vs[i*graphics.VertexFloatCount+11] = v.Custom3
 	}
 
 	is := i.ensureTmpIndices(len(indices))

internal/builtinshader/shader.go

@@ -170,7 +170,7 @@ var ScreenShaderSource = []byte(`//kage:unit pixels
 
 package main
 
-func Fragment(dstPos vec4, srcPos vec2, color vec4) vec4 {
+func Fragment(dstPos vec4, srcPos vec2) vec4 {
 	// Blend source colors in a square region, which size is 1/scale.
 	scale := imageDstSize()/imageSrc0Size()
 	pos := srcPos
@@ -193,7 +193,7 @@ var ClearShaderSource = []byte(`//kage:unit pixels
 
 package main
 
-func Fragment(dstPos vec4, srcPos vec2, color vec4) vec4 {
+func Fragment() vec4 {
 	return vec4(0)
 }
 `)

internal/graphics/shader.go

@@ -154,8 +154,8 @@ func imageSrc%[1]dAt(pos vec2) vec4 {
 	shaderSuffix += `
 var __projectionMatrix mat4
 
-func __vertex(dstPos vec2, srcPos vec2, color vec4) (vec4, vec2, vec4) {
-	return __projectionMatrix * vec4(dstPos, 0, 1), srcPos, color
+func __vertex(dstPos vec2, srcPos vec2, color vec4, custom vec4) (vec4, vec2, vec4, vec4) {
+	return __projectionMatrix * vec4(dstPos, 0, 1), srcPos, color, custom
 }
 `
 	return shaderSuffix, nil

internal/graphics/vertex.go

@@ -39,7 +39,7 @@ const (
 )
 
 const (
-	VertexFloatCount = 8
+	VertexFloatCount = 12
 )
 
 var (

internal/graphicsdriver/directx/graphics11_windows.go

@@ -27,7 +27,10 @@ import (
 	"github.com/hajimehoshi/ebiten/v2/internal/shaderir/hlsl"
 )
 
-var inputElementDescsForDX11 = []_D3D11_INPUT_ELEMENT_DESC{
+var inputElementDescsForDX11 []_D3D11_INPUT_ELEMENT_DESC
+
+func init() {
+	inputElementDescsForDX11 = []_D3D11_INPUT_ELEMENT_DESC{
 		{
 			SemanticName:         &([]byte("POSITION\000"))[0],
 			SemanticIndex:        0,
@@ -55,6 +58,25 @@ var inputElementDescsForDX11 = []_D3D11_INPUT_ELEMENT_DESC{
 			InputSlotClass:       _D3D11_INPUT_PER_VERTEX_DATA,
 			InstanceDataStepRate: 0,
 		},
+	}
+	diff := graphics.VertexFloatCount - 8
+	if diff == 0 {
+		return
+	}
+	if diff%4 != 0 {
+		panic("directx: unexpected attribute layout")
+	}
+	for i := 0; i < diff/4; i++ {
+		inputElementDescsForDX11 = append(inputElementDescsForDX11, _D3D11_INPUT_ELEMENT_DESC{
+			SemanticName:         &([]byte("COLOR\000"))[0],
+			SemanticIndex:        uint32(i) + 1,
+			Format:               _DXGI_FORMAT_R32G32B32A32_FLOAT,
+			InputSlot:            0,
+			AlignedByteOffset:    _D3D11_APPEND_ALIGNED_ELEMENT,
+			InputSlotClass:       _D3D11_INPUT_PER_VERTEX_DATA,
+			InstanceDataStepRate: 0,
+		})
+	}
 }
 
 func blendFactorToBlend11(f graphicsdriver.BlendFactor, alpha bool) _D3D11_BLEND {

internal/graphicsdriver/directx/pipeline12_windows.go

@@ -23,7 +23,10 @@ import (
 	"github.com/hajimehoshi/ebiten/v2/internal/graphicsdriver"
 )
 
-var inputElementDescsForDX12 = []_D3D12_INPUT_ELEMENT_DESC{
+var inputElementDescsForDX12 []_D3D12_INPUT_ELEMENT_DESC
+
+func init() {
+	inputElementDescsForDX12 = []_D3D12_INPUT_ELEMENT_DESC{
 		{
 			SemanticName:         &([]byte("POSITION\000"))[0],
 			SemanticIndex:        0,
@@ -51,6 +54,25 @@ var inputElementDescsForDX12 = []_D3D12_INPUT_ELEMENT_DESC{
 			InputSlotClass:       _D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA,
 			InstanceDataStepRate: 0,
 		},
+	}
+	diff := graphics.VertexFloatCount - 8
+	if diff == 0 {
+		return
+	}
+	if diff%4 != 0 {
+		panic("directx: unexpected attribute layout")
+	}
+	for i := 0; i < diff/4; i++ {
+		inputElementDescsForDX12 = append(inputElementDescsForDX12, _D3D12_INPUT_ELEMENT_DESC{
+			SemanticName:         &([]byte("COLOR\000"))[0],
+			SemanticIndex:        uint32(i) + 1,
+			Format:               _DXGI_FORMAT_R32G32B32A32_FLOAT,
+			InputSlot:            0,
+			AlignedByteOffset:    _D3D12_APPEND_ALIGNED_ELEMENT,
+			InputSlotClass:       _D3D12_INPUT_CLASSIFICATION_PER_VERTEX_DATA,
+			InstanceDataStepRate: 0,
+		})
+	}
 }
 
 const numDescriptorsPerFrame = 32

internal/graphicsdriver/directx/shader_windows.go

@@ -102,7 +102,7 @@ func compileShader(program *shaderir.Program) (vsh, psh *_ID3DBlob, ferr error)
 		return vsh, psh, nil
 	}
 
-	vs, ps := hlsl.Compile(program)
+	vs, ps, _ := hlsl.Compile(program)
 	var flag uint32 = uint32(_D3DCOMPILE_OPTIMIZATION_LEVEL3)
 
 	var wg errgroup.Group

internal/graphicsdriver/opengl/program.go

@@ -114,17 +114,17 @@ func init() {
 		},
 	}
 	n := theArrayBufferLayout.float32Count()
-	if n > graphics.VertexFloatCount {
-		panic("opengl: the array buffer layout is too large")
+	diff := graphics.VertexFloatCount - n
+	if diff == 0 {
+		return
 	}
-	if n < graphics.VertexFloatCount {
-		d := graphics.VertexFloatCount - n
-		if d > 4 {
-			panic("opengl: the array buffer layout is too small")
+	if diff%4 != 0 {
+		panic("opengl: unexpected attribute layout")
 	}
+	for i := 0; i < diff/4; i++ {
 		theArrayBufferLayout.addPart(arrayBufferLayoutPart{
-			name: "A3",
-			num:  d,
+			name: fmt.Sprintf("A%d", i+3),
+			num:  4,
 		})
 	}
 }

internal/shader/shader.go

@@ -290,10 +290,10 @@ func (cs *compileState) parse(f *ast.File) {
 
 	// Parse function names so that any other function call the others.
 	// The function data is provisional and will be updated soon.
-	var vertexInParams []shaderir.Type
-	var vertexOutParams []shaderir.Type
-	var fragmentInParams []shaderir.Type
-	var fragmentOutParams []shaderir.Type
+	var vertexInParams []variable
+	var vertexOutParams []variable
+	var fragmentInParams []variable
+	var fragmentOutParams []variable
 	var fragmentReturnType shaderir.Type
 	for _, d := range f.Decls {
 		fd, ok := d.(*ast.FuncDecl)
@@ -310,6 +310,19 @@ func (cs *compileState) parse(f *ast.File) {
 		}
 
 		inParams, outParams, ret := cs.parseFuncParams(&cs.global, n, fd)
+
+		if n == cs.vertexEntry {
+			vertexInParams = inParams
+			vertexOutParams = outParams
+			continue
+		}
+		if n == cs.fragmentEntry {
+			fragmentInParams = inParams
+			fragmentOutParams = outParams
+			fragmentReturnType = ret
+			continue
+		}
+
 		var inT, outT []shaderir.Type
 		for _, v := range inParams {
 			inT = append(inT, v.typ)
@@ -317,19 +330,6 @@ func (cs *compileState) parse(f *ast.File) {
 		for _, v := range outParams {
 			outT = append(outT, v.typ)
 		}
-
-		if n == cs.vertexEntry {
-			vertexInParams = inT
-			vertexOutParams = outT
-			continue
-		}
-		if n == cs.fragmentEntry {
-			fragmentInParams = inT
-			fragmentOutParams = outT
-			fragmentReturnType = ret
-			continue
-		}
-
 		cs.funcs = append(cs.funcs, function{
 			name: n,
 			ir: shaderir.Func{
@@ -345,25 +345,21 @@ func (cs *compileState) parse(f *ast.File) {
 	// Check varying variables.
 	// In testings, there might not be vertex and fragment entry points.
 	if len(vertexOutParams) > 0 && len(fragmentInParams) > 0 {
-		if len(vertexOutParams) != len(fragmentInParams) {
-			cs.addError(0, "the number of vertex entry point's returning values and the number of fragment entry point's params must be the same")
+		for i, p := range vertexOutParams {
+			if len(fragmentInParams) <= i {
+				break
 			}
-		for i, t := range vertexOutParams {
-			if !t.Equal(&fragmentInParams[i]) {
-				cs.addError(0, "vertex entry point's returning value types and fragment entry point's param types must match")
+			t := fragmentInParams[i].typ
+			if !p.typ.Equal(&t) {
+				name := fragmentInParams[i].name
+				cs.addError(0, fmt.Sprintf("fragment argument %s must be %s but was %s", name, p.typ.String(), t.String()))
 			}
 		}
 
 		// The first out-param is treated as gl_Position in GLSL.
-		if vertexOutParams[0].Main != shaderir.Vec4 {
+		if vertexOutParams[0].typ.Main != shaderir.Vec4 {
 			cs.addError(0, "vertex entry point must have at least one returning vec4 value for a position")
 		}
-		if len(fragmentInParams) == 0 {
-			cs.addError(0, "fragment entry point must have at least one vec4 parameter for a position")
-		}
-		if fragmentInParams[0].Main != shaderir.Vec4 {
-			cs.addError(0, "fragment entry point must have at least one vec4 parameter for a position")
-		}
 		if len(fragmentOutParams) != 0 || fragmentReturnType.Main != shaderir.Vec4 {
 			cs.addError(0, "fragment entry point must have one returning vec4 value for a color")
 		}
@@ -373,12 +369,16 @@ func (cs *compileState) parse(f *ast.File) {
 		return
 	}
 
-	// Set attribute varying veraibles.
-	cs.ir.Attributes = append(cs.ir.Attributes, vertexInParams...)
+	// Set attribute and varying veraibles.
+	for _, p := range vertexInParams {
+		cs.ir.Attributes = append(cs.ir.Attributes, p.typ)
+	}
 	if len(vertexOutParams) > 0 {
 		// TODO: Check that these params are not arrays or structs
 		// The 0th argument is a special variable for position and is not included in varying variables.
-		cs.ir.Varyings = append(cs.ir.Varyings, vertexOutParams[1:]...)
+		for _, p := range vertexOutParams[1:] {
+			cs.ir.Varyings = append(cs.ir.Varyings, p.typ)
+		}
 	}
 
 	// Parse functions.
@@ -820,6 +820,25 @@ func (cs *compileState) parseFunc(block *block, d *ast.FuncDecl) (function, bool
 	}
 
 	inParams, outParams, returnType := cs.parseFuncParams(block, d.Name.Name, d)
+	if d.Name.Name == cs.fragmentEntry {
+		if len(inParams) == 0 {
+			inParams = append(inParams, variable{
+				name: "_",
+				typ:  shaderir.Type{Main: shaderir.Vec4},
+			})
+		}
+		// The 0th inParams is a special variable for position and is not included in varying variables.
+		if diff := len(cs.ir.Varyings) - (len(inParams) - 1); diff > 0 {
+			// inParams is not enough when the vertex shader has more returning values than the fragment shader's arguments.
+			orig := len(inParams) - 1
+			for i := 0; i < diff; i++ {
+				inParams = append(inParams, variable{
+					name: "_",
+					typ:  cs.ir.Varyings[orig+i],
+				})
+			}
+		}
+	}
 	b, ok := cs.parseBlock(block, d.Name.Name, d.Body.List, inParams, outParams, returnType, true)
 	if !ok {
 		return function{}, false

internal/shader/shader_test.go

@@ -31,32 +31,24 @@ import (
 
 func glslVertexNormalize(str string) string {
 	p := glsl.VertexPrelude(glsl.GLSLVersionDefault)
-	if strings.HasPrefix(str, p) {
-		str = str[len(p):]
-	}
+	str = strings.TrimPrefix(str, p)
 	return strings.TrimSpace(str)
 }
 
 func glslFragmentNormalize(str string) string {
 	p := glsl.FragmentPrelude(glsl.GLSLVersionDefault)
-	if strings.HasPrefix(str, p) {
-		str = str[len(p):]
-	}
+	str = strings.TrimPrefix(str, p)
 	return strings.TrimSpace(str)
 }
 
-func hlslNormalize(str string) string {
-	if strings.HasPrefix(str, hlsl.Prelude) {
-		str = str[len(hlsl.Prelude):]
-	}
+func hlslNormalize(str string, prelude string) string {
+	str = strings.TrimPrefix(str, prelude)
 	return strings.TrimSpace(str)
 }
 
 func metalNormalize(str string) string {
 	prelude := msl.Prelude(shaderir.Texels)
-	if strings.HasPrefix(str, prelude) {
-		str = str[len(prelude):]
-	}
+	str = strings.TrimPrefix(str, prelude)
 	return strings.TrimSpace(str)
 }
 
@@ -188,8 +180,8 @@ func TestCompile(t *testing.T) {
 			}
 
 			if tc.HLSL != nil {
-				vs, _ := hlsl.Compile(s)
-				if got, want := hlslNormalize(vs), hlslNormalize(string(tc.HLSL)); got != want {
+				vs, _, prelude := hlsl.Compile(s)
+				if got, want := hlslNormalize(vs, prelude), hlslNormalize(string(tc.HLSL), prelude); got != want {
 					compare(t, "HLSL", got, want)
 				}
 			}

internal/shader/testdata/vertex.expected.hlsl

@@ -2,7 +2,7 @@ cbuffer Uniforms : register(b0) {
 	float2 U0 : packoffset(c0);
 }
 
-Varyings VSMain(float2 A0 : POSITION, float2 A1 : TEXCOORD, float4 A2 : COLOR) {
+Varyings VSMain(float2 A0 : POSITION, float2 A1 : TEXCOORD, float4 A2 : COLOR0) {
 	Varyings varyings;
 	float4x4 l0 = 0.0;
 	varyings.Position = 0.0;

internal/shaderir/hlsl/hlsl.go

@@ -52,13 +52,7 @@ func (c *compileContext) structName(p *shaderir.Program, t *shaderir.Type) strin
 	return n
 }
 
-const Prelude = `struct Varyings {
-	float4 Position : SV_POSITION;
-	float2 M0 : TEXCOORD0;
-	float4 M1 : COLOR;
-};
-
-float mod(float x, float y) {
+const utilFuncs = `float mod(float x, float y) {
 	return x - y * floor(x/y);
 }
 
@@ -86,7 +80,7 @@ float4x4 float4x4FromScalar(float x) {
 	return float4x4(x, 0, 0, 0, 0, x, 0, 0, 0, 0, x, 0, 0, 0, 0, x);
 }`
 
-func Compile(p *shaderir.Program) (vertexShader, pixelShader string) {
+func Compile(p *shaderir.Program) (vertexShader, pixelShader, prelude string) {
 	offsets := CalcUniformMemoryOffsets(p)
 
 	c := &compileContext{
@@ -94,7 +88,30 @@ func Compile(p *shaderir.Program) (vertexShader, pixelShader string) {
 	}
 
 	var lines []string
-	lines = append(lines, strings.Split(Prelude, "\n")...)
+	lines = append(lines, strings.Split(utilFuncs, "\n")...)
+
+	lines = append(lines, "", "struct Varyings {")
+	lines = append(lines, "\tfloat4 Position : SV_POSITION;")
+	if len(p.Varyings) > 0 {
+		for i, v := range p.Varyings {
+			switch i {
+			case 0:
+				lines = append(lines, fmt.Sprintf("\tfloat2 M%d : TEXCOORD;", i))
+			case 1:
+				lines = append(lines, fmt.Sprintf("\tfloat4 M%d : COLOR0;", i))
+			default:
+				// Use COLOR[n] as a general purpose varying.
+				if v.Main != shaderir.Vec4 {
+					lines = append(lines, fmt.Sprintf("\t?(unexpected type: %s) M%d : COLOR%d;", v, i, i-1))
+				} else {
+					lines = append(lines, fmt.Sprintf("\tfloat4 M%d : COLOR%d;", i, i-1))
+				}
+			}
+		}
+	}
+	lines = append(lines, "};")
+	prelude = strings.Join(lines, "\n")
+
 	lines = append(lines, "", "{{.Structs}}")
 
 	if len(p.Uniforms) > 0 {
@@ -156,7 +173,25 @@ func Compile(p *shaderir.Program) (vertexShader, pixelShader string) {
 	}
 	if p.VertexFunc.Block != nil && len(p.VertexFunc.Block.Stmts) > 0 {
 		vslines = append(vslines, "")
-		vslines = append(vslines, "Varyings VSMain(float2 A0 : POSITION, float2 A1 : TEXCOORD, float4 A2 : COLOR) {")
+		var args []string
+		for i, a := range p.Attributes {
+			switch i {
+			case 0:
+				args = append(args, fmt.Sprintf("float2 A%d : POSITION", i))
+			case 1:
+				args = append(args, fmt.Sprintf("float2 A%d : TEXCOORD", i))
+			case 2:
+				args = append(args, fmt.Sprintf("float4 A%d : COLOR0", i))
+			default:
+				// Use COLOR[n] as a general purpose varying.
+				if a.Main != shaderir.Vec4 {
+					args = append(args, fmt.Sprintf("?(unexpected type: %s) A%d : COLOR%d", a, i, i-2))
+				} else {
+					args = append(args, fmt.Sprintf("float4 A%d : COLOR%d", i, i-2))
+				}
+			}
+		}
+		vslines = append(vslines, "Varyings VSMain("+strings.Join(args, ", ")+") {")
 		vslines = append(vslines, fmt.Sprintf("\tVaryings %s;", vsOut))
 		vslines = append(vslines, c.block(p, p.VertexFunc.Block, p.VertexFunc.Block, 0)...)
 		if last := fmt.Sprintf("\treturn %s;", vsOut); vslines[len(vslines)-1] != last {

shader_test.go

@@ -2595,3 +2595,193 @@ func Fragment(dstPos vec4, srcPos vec2, color vec4) vec4 {
 		}
 	}
 }
+
+func TestShaderCustomValues(t *testing.T) {
+	const w, h = 16, 16
+
+	dst := ebiten.NewImage(w, h)
+	s, err := ebiten.NewShader([]byte(`//kage:unit pixels
+
+package main
+
+func Fragment(dstPos vec4, srcPos vec2, color vec4, custom vec4) vec4 {
+	return custom
+}
+`))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	clr := color.RGBA{R: 0x10, G: 0x20, B: 0x30, A: 0x40}
+	dst.DrawTrianglesShader([]ebiten.Vertex{
+		{
+			DstX:    0,
+			DstY:    0,
+			SrcX:    0,
+			SrcY:    0,
+			ColorR:  1,
+			ColorG:  1,
+			ColorB:  1,
+			ColorA:  1,
+			Custom0: float32(clr.R) / 0xff,
+			Custom1: float32(clr.G) / 0xff,
+			Custom2: float32(clr.B) / 0xff,
+			Custom3: float32(clr.A) / 0xff,
+		},
+		{
+			DstX:    w,
+			DstY:    0,
+			SrcX:    w,
+			SrcY:    0,
+			ColorR:  1,
+			ColorG:  1,
+			ColorB:  1,
+			ColorA:  1,
+			Custom0: float32(clr.R) / 0xff,
+			Custom1: float32(clr.G) / 0xff,
+			Custom2: float32(clr.B) / 0xff,
+			Custom3: float32(clr.A) / 0xff,
+		},
+		{
+			DstX:    0,
+			DstY:    h,
+			SrcX:    0,
+			SrcY:    h,
+			ColorR:  1,
+			ColorG:  1,
+			ColorB:  1,
+			ColorA:  1,
+			Custom0: float32(clr.R) / 0xff,
+			Custom1: float32(clr.G) / 0xff,
+			Custom2: float32(clr.B) / 0xff,
+			Custom3: float32(clr.A) / 0xff,
+		},
+		{
+			DstX:    w,
+			DstY:    h,
+			SrcX:    w,
+			SrcY:    h,
+			ColorR:  1,
+			ColorG:  1,
+			ColorB:  1,
+			ColorA:  1,
+			Custom0: float32(clr.R) / 0xff,
+			Custom1: float32(clr.G) / 0xff,
+			Custom2: float32(clr.B) / 0xff,
+			Custom3: float32(clr.A) / 0xff,
+		},
+	}, []uint16{0, 1, 2, 1, 2, 3}, s, nil)
+
+	for j := 0; j < h; j++ {
+		for i := 0; i < w; i++ {
+			got := dst.At(i, j).(color.RGBA)
+			want := clr
+			if !sameColors(got, want, 2) {
+				t.Errorf("dst.At(%d, %d): got: %v, want: %v", i, j, got, want)
+			}
+		}
+	}
+}
+
+func TestShaderFragmentLessArguments(t *testing.T) {
+	const w, h = 16, 16
+
+	s0, err := ebiten.NewShader([]byte(`//kage:unit pixels
+
+package main
+
+func Fragment() vec4 {
+	return vec4(1, 0, 0, 1)
+}
+`))
+	if err != nil {
+		t.Fatal(err)
+	}
+	s1, err := ebiten.NewShader([]byte(`//kage:unit pixels
+
+package main
+
+func Fragment(dstPos vec4) vec4 {
+	return vec4(0, 1, 0, 1)
+}
+`))
+	if err != nil {
+		t.Fatal(err)
+	}
+	s2, err := ebiten.NewShader([]byte(`//kage:unit pixels
+
+package main
+
+func Fragment(dstPos vec4, srcPos vec2) vec4 {
+	return vec4(0, 0, 1, 1)
+}
+`))
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	dst := ebiten.NewImage(w, h)
+	for idx, s := range []*ebiten.Shader{s0, s1, s2} {
+		dst.Clear()
+		dst.DrawTrianglesShader([]ebiten.Vertex{
+			{
+				DstX:   0,
+				DstY:   0,
+				SrcX:   0,
+				SrcY:   0,
+				ColorR: 1,
+				ColorG: 1,
+				ColorB: 1,
+				ColorA: 1,
+			},
+			{
+				DstX:   w,
+				DstY:   0,
+				SrcX:   w,
+				SrcY:   0,
+				ColorR: 1,
+				ColorG: 1,
+				ColorB: 1,
+				ColorA: 1,
+			},
+			{
+				DstX:   0,
+				DstY:   h,
+				SrcX:   0,
+				SrcY:   h,
+				ColorR: 1,
+				ColorG: 1,
+				ColorB: 1,
+				ColorA: 1,
+			},
+			{
+				DstX:   w,
+				DstY:   h,
+				SrcX:   w,
+				SrcY:   h,
+				ColorR: 1,
+				ColorG: 1,
+				ColorB: 1,
+				ColorA: 1,
+			},
+		}, []uint16{0, 1, 2, 1, 2, 3}, s, nil)
+
+		for j := 0; j < h; j++ {
+			for i := 0; i < w; i++ {
+				got := dst.At(i, j).(color.RGBA)
+				var want color.RGBA
+				switch idx {
+				case 0:
+					want = color.RGBA{R: 0xff, A: 0xff}
+				case 1:
+					want = color.RGBA{G: 0xff, A: 0xff}
+				case 2:
+					want = color.RGBA{B: 0xff, A: 0xff}
+				}
+				if !sameColors(got, want, 2) {
+					t.Errorf("dst.At(%d, %d): got: %v, want: %v", i, j, got, want)
+				}
+			}
+		}
+	}
+}