diff --git a/internal/cgen/base/pixconv-submodule.c b/internal/cgen/base/pixconv-submodule.c
index 50cb427..0336408 100644
--- a/internal/cgen/base/pixconv-submodule.c
+++ b/internal/cgen/base/pixconv-submodule.c
@@ -1887,7 +1887,6 @@
     d[0] = s0;
     d[1] = s1;
     d[2] = s2;
-    d[3] = 0xFF;
 
     s += 1 * 4;
     d += 1 * 3;
@@ -2041,7 +2040,6 @@
     d[0] = s2;
     d[1] = s1;
     d[2] = s0;
-    d[3] = 0xFF;
 
     s += 1 * 4;
     d += 1 * 3;
diff --git a/internal/cgen/data/data.go b/internal/cgen/data/data.go
index 20cf7aa..da6e086 100644
--- a/internal/cgen/data/data.go
+++ b/internal/cgen/data/data.go
@@ -620,13 +620,13 @@
 	"t n = len;\n\n  // TODO: unroll.\n\n  while (n >= 1) {\n    // Extract 16-bit color components.\n    uint32_t sa = 0x101 * ((uint32_t)s[3]);\n    uint32_t sr = 0x101 * ((uint32_t)s[2]);\n    uint32_t sg = 0x101 * ((uint32_t)s[1]);\n    uint32_t sb = 0x101 * ((uint32_t)s[0]);\n    uint32_t dr = 0x101 * ((uint32_t)d[2]);\n    uint32_t dg = 0x101 * ((uint32_t)d[1]);\n    uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n    // Calculate the inverse of the src-alpha: how much of the dst to keep.\n    uint32_t ia = 0xFFFF - sa;\n\n    // Composite src (nonpremul) over dst (premul).\n    dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n    dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n    db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n    // Convert from 16-bit color to 8-bit color.\n    d[0] = (uint8_t)(db >> 8);\n    d[1] = (uint8_t)(dg >> 8);\n    d[2] = (uint8_t)(dr >> 8);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__bgra_nonpremul_4x16le__src_over(\n    uint8_t* dst_ptr,\n    size_t dst_le" +
 	"n,\n    uint8_t* dst_palette_ptr,\n    size_t dst_palette_len,\n    const uint8_t* src_ptr,\n    size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len8 = src_len / 8;\n  size_t len = (dst_len3 < src_len8) ? dst_len3 : src_len8;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  // TODO: unroll.\n\n  while (n >= 1) {\n    // Extract 16-bit color components.\n    uint32_t sa = ((uint32_t)wuffs_base__peek_u16le__no_bounds_check(s + 6));\n    uint32_t sr = ((uint32_t)wuffs_base__peek_u16le__no_bounds_check(s + 4));\n    uint32_t sg = ((uint32_t)wuffs_base__peek_u16le__no_bounds_check(s + 2));\n    uint32_t sb = ((uint32_t)wuffs_base__peek_u16le__no_bounds_check(s + 0));\n    uint32_t dr = 0x101 * ((uint32_t)d[2]);\n    uint32_t dg = 0x101 * ((uint32_t)d[1]);\n    uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n    // Calculate the inverse of the src-alpha: how much of the dst to keep.\n    uint32_t ia = 0xFFFF - sa;\n\n    // Composite src (nonpremul) over dst (premul).\n    dr = ((sr * sa) + (dr * ia" +
 	")) / 0xFFFF;\n    dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n    db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n    // Convert from 16-bit color to 8-bit color.\n    d[0] = (uint8_t)(db >> 8);\n    d[1] = (uint8_t)(dg >> 8);\n    d[2] = (uint8_t)(dr >> 8);\n\n    s += 1 * 8;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__bgra_premul__src(uint8_t* dst_ptr,\n                                                  size_t dst_len,\n                                                  uint8_t* dst_palette_ptr,\n                                                  size_t dst_palette_len,\n                                                  const uint8_t* src_ptr,\n                                                  size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    uint8_t s0 = s[0];\n    uint8_t s1 = s[1];\n    u" +
-	"int8_t s2 = s[2];\n    d[0] = s0;\n    d[1] = s1;\n    d[2] = s2;\n    d[3] = 0xFF;\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__bgra_premul__src_over(uint8_t* dst_ptr,\n                                                       size_t dst_len,\n                                                       uint8_t* dst_palette_ptr,\n                                                       size_t dst_palette_len,\n                                                       const uint8_t* src_ptr,\n                                                       size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    // Extract 16-bit color components.\n    uint32_t sa = 0x101 * ((uint32_t)s[3]);\n    uint32_t sr = 0x101 * ((uint32_t)s[2]);\n    uint32_t sg = 0x101 * ((uint32_t)s[1]);\n    uint32_" +
-	"t sb = 0x101 * ((uint32_t)s[0]);\n    uint32_t dr = 0x101 * ((uint32_t)d[2]);\n    uint32_t dg = 0x101 * ((uint32_t)d[1]);\n    uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n    // Calculate the inverse of the src-alpha: how much of the dst to keep.\n    uint32_t ia = 0xFFFF - sa;\n\n    // Composite src (premul) over dst (premul).\n    dr = sr + ((dr * ia) / 0xFFFF);\n    dg = sg + ((dg * ia) / 0xFFFF);\n    db = sb + ((db * ia) / 0xFFFF);\n\n    // Convert from 16-bit color to 8-bit color.\n    d[0] = (uint8_t)(db >> 8);\n    d[1] = (uint8_t)(dg >> 8);\n    d[2] = (uint8_t)(dr >> 8);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__rgba_nonpremul__src(uint8_t* dst_ptr,\n                                                     size_t dst_len,\n                                                     uint8_t* dst_palette_ptr,\n                                                     size_t dst_palette_len,\n                                                     const uint8_t*" +
-	" src_ptr,\n                                                     size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  // TODO: unroll.\n\n  while (n >= 1) {\n    uint32_t s0 = wuffs_base__swap_u32_argb_abgr(\n        wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n            wuffs_base__peek_u32le__no_bounds_check(s + (0 * 4))));\n    wuffs_base__poke_u24le__no_bounds_check(d + (0 * 3), s0);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__rgba_nonpremul__src_over(\n    uint8_t* dst_ptr,\n    size_t dst_len,\n    uint8_t* dst_palette_ptr,\n    size_t dst_palette_len,\n    const uint8_t* src_ptr,\n    size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  cons" +
-	"t uint8_t* s = src_ptr;\n  size_t n = len;\n\n  // TODO: unroll.\n\n  while (n >= 1) {\n    // Extract 16-bit color components.\n    uint32_t sa = 0x101 * ((uint32_t)s[3]);\n    uint32_t sb = 0x101 * ((uint32_t)s[2]);\n    uint32_t sg = 0x101 * ((uint32_t)s[1]);\n    uint32_t sr = 0x101 * ((uint32_t)s[0]);\n    uint32_t dr = 0x101 * ((uint32_t)d[2]);\n    uint32_t dg = 0x101 * ((uint32_t)d[1]);\n    uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n    // Calculate the inverse of the src-alpha: how much of the dst to keep.\n    uint32_t ia = 0xFFFF - sa;\n\n    // Composite src (nonpremul) over dst (premul).\n    dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n    dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n    db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n    // Convert from 16-bit color to 8-bit color.\n    d[0] = (uint8_t)(db >> 8);\n    d[1] = (uint8_t)(dg >> 8);\n    d[2] = (uint8_t)(dr >> 8);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__rgba_premul__src(uint8_t* dst_ptr,\n      " +
-	"                                            size_t dst_len,\n                                                  uint8_t* dst_palette_ptr,\n                                                  size_t dst_palette_len,\n                                                  const uint8_t* src_ptr,\n                                                  size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    uint8_t s0 = s[0];\n    uint8_t s1 = s[1];\n    uint8_t s2 = s[2];\n    d[0] = s2;\n    d[1] = s1;\n    d[2] = s0;\n    d[3] = 0xFF;\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__rgba_premul__src_over(uint8_t* dst_ptr,\n                                                       size_t dst_len,\n                                                       uint8_t* dst_palette_ptr,\n           " +
-	"                                            size_t dst_palette_len,\n                                                       const uint8_t* src_ptr,\n                                                       size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    // Extract 16-bit color components.\n    uint32_t sa = 0x101 * ((uint32_t)s[3]);\n    uint32_t sb = 0x101 * ((uint32_t)s[2]);\n    uint32_t sg = 0x101 * ((uint32_t)s[1]);\n    uint32_t sr = 0x101 * ((uint32_t)s[0]);\n    uint32_t dr = 0x101 * ((uint32_t)d[2]);\n    uint32_t dg = 0x101 * ((uint32_t)d[1]);\n    uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n    // Calculate the inverse of the src-alpha: how much of the dst to keep.\n    uint32_t ia = 0xFFFF - sa;\n\n    // Composite src (premul) over dst (premul).\n    dr = sr + ((dr * ia) / 0xFFFF);\n    dg = sg + ((dg * ia) / 0xFFFF);\n    db =" +
-	" sb + ((db * ia) / 0xFFFF);\n\n    // Convert from 16-bit color to 8-bit color.\n    d[0] = (uint8_t)(db >> 8);\n    d[1] = (uint8_t)(dg >> 8);\n    d[2] = (uint8_t)(dr >> 8);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\n" +
+	"int8_t s2 = s[2];\n    d[0] = s0;\n    d[1] = s1;\n    d[2] = s2;\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__bgra_premul__src_over(uint8_t* dst_ptr,\n                                                       size_t dst_len,\n                                                       uint8_t* dst_palette_ptr,\n                                                       size_t dst_palette_len,\n                                                       const uint8_t* src_ptr,\n                                                       size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    // Extract 16-bit color components.\n    uint32_t sa = 0x101 * ((uint32_t)s[3]);\n    uint32_t sr = 0x101 * ((uint32_t)s[2]);\n    uint32_t sg = 0x101 * ((uint32_t)s[1]);\n    uint32_t sb = 0x101 * ((" +
+	"uint32_t)s[0]);\n    uint32_t dr = 0x101 * ((uint32_t)d[2]);\n    uint32_t dg = 0x101 * ((uint32_t)d[1]);\n    uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n    // Calculate the inverse of the src-alpha: how much of the dst to keep.\n    uint32_t ia = 0xFFFF - sa;\n\n    // Composite src (premul) over dst (premul).\n    dr = sr + ((dr * ia) / 0xFFFF);\n    dg = sg + ((dg * ia) / 0xFFFF);\n    db = sb + ((db * ia) / 0xFFFF);\n\n    // Convert from 16-bit color to 8-bit color.\n    d[0] = (uint8_t)(db >> 8);\n    d[1] = (uint8_t)(dg >> 8);\n    d[2] = (uint8_t)(dr >> 8);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__rgba_nonpremul__src(uint8_t* dst_ptr,\n                                                     size_t dst_len,\n                                                     uint8_t* dst_palette_ptr,\n                                                     size_t dst_palette_len,\n                                                     const uint8_t* src_ptr,\n       " +
+	"                                              size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  // TODO: unroll.\n\n  while (n >= 1) {\n    uint32_t s0 = wuffs_base__swap_u32_argb_abgr(\n        wuffs_base__color_u32_argb_nonpremul__as__color_u32_argb_premul(\n            wuffs_base__peek_u32le__no_bounds_check(s + (0 * 4))));\n    wuffs_base__poke_u24le__no_bounds_check(d + (0 * 3), s0);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__rgba_nonpremul__src_over(\n    uint8_t* dst_ptr,\n    size_t dst_len,\n    uint8_t* dst_palette_ptr,\n    size_t dst_palette_len,\n    const uint8_t* src_ptr,\n    size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = sr" +
+	"c_ptr;\n  size_t n = len;\n\n  // TODO: unroll.\n\n  while (n >= 1) {\n    // Extract 16-bit color components.\n    uint32_t sa = 0x101 * ((uint32_t)s[3]);\n    uint32_t sb = 0x101 * ((uint32_t)s[2]);\n    uint32_t sg = 0x101 * ((uint32_t)s[1]);\n    uint32_t sr = 0x101 * ((uint32_t)s[0]);\n    uint32_t dr = 0x101 * ((uint32_t)d[2]);\n    uint32_t dg = 0x101 * ((uint32_t)d[1]);\n    uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n    // Calculate the inverse of the src-alpha: how much of the dst to keep.\n    uint32_t ia = 0xFFFF - sa;\n\n    // Composite src (nonpremul) over dst (premul).\n    dr = ((sr * sa) + (dr * ia)) / 0xFFFF;\n    dg = ((sg * sa) + (dg * ia)) / 0xFFFF;\n    db = ((sb * sa) + (db * ia)) / 0xFFFF;\n\n    // Convert from 16-bit color to 8-bit color.\n    d[0] = (uint8_t)(db >> 8);\n    d[1] = (uint8_t)(dg >> 8);\n    d[2] = (uint8_t)(dr >> 8);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__rgba_premul__src(uint8_t* dst_ptr,\n                       " +
+	"                           size_t dst_len,\n                                                  uint8_t* dst_palette_ptr,\n                                                  size_t dst_palette_len,\n                                                  const uint8_t* src_ptr,\n                                                  size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    uint8_t s0 = s[0];\n    uint8_t s1 = s[1];\n    uint8_t s2 = s[2];\n    d[0] = s2;\n    d[1] = s1;\n    d[2] = s0;\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgr__rgba_premul__src_over(uint8_t* dst_ptr,\n                                                       size_t dst_len,\n                                                       uint8_t* dst_palette_ptr,\n                                             " +
+	"          size_t dst_palette_len,\n                                                       const uint8_t* src_ptr,\n                                                       size_t src_len) {\n  size_t dst_len3 = dst_len / 3;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len3 < src_len4) ? dst_len3 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    // Extract 16-bit color components.\n    uint32_t sa = 0x101 * ((uint32_t)s[3]);\n    uint32_t sb = 0x101 * ((uint32_t)s[2]);\n    uint32_t sg = 0x101 * ((uint32_t)s[1]);\n    uint32_t sr = 0x101 * ((uint32_t)s[0]);\n    uint32_t dr = 0x101 * ((uint32_t)d[2]);\n    uint32_t dg = 0x101 * ((uint32_t)d[1]);\n    uint32_t db = 0x101 * ((uint32_t)d[0]);\n\n    // Calculate the inverse of the src-alpha: how much of the dst to keep.\n    uint32_t ia = 0xFFFF - sa;\n\n    // Composite src (premul) over dst (premul).\n    dr = sr + ((dr * ia) / 0xFFFF);\n    dg = sg + ((dg * ia) / 0xFFFF);\n    db = sb + ((db * ia) / 0xFFFF);\n\n    /" +
+	"/ Convert from 16-bit color to 8-bit color.\n    d[0] = (uint8_t)(db >> 8);\n    d[1] = (uint8_t)(dg >> 8);\n    d[2] = (uint8_t)(dr >> 8);\n\n    s += 1 * 4;\n    d += 1 * 3;\n    n -= 1;\n  }\n\n  return len;\n}\n\n" +
 	"" +
 	"// --------\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul__src_over(\n    uint8_t* dst_ptr,\n    size_t dst_len,\n    uint8_t* dst_palette_ptr,\n    size_t dst_palette_len,\n    const uint8_t* src_ptr,\n    size_t src_len) {\n  size_t dst_len4 = dst_len / 4;\n  size_t src_len4 = src_len / 4;\n  size_t len = (dst_len4 < src_len4) ? dst_len4 : src_len4;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    uint32_t d0 = wuffs_base__peek_u32le__no_bounds_check(d + (0 * 4));\n    uint32_t s0 = wuffs_base__peek_u32le__no_bounds_check(s + (0 * 4));\n    wuffs_base__poke_u32le__no_bounds_check(\n        d + (0 * 4),\n        wuffs_base__composite_nonpremul_nonpremul_u32_axxx(d0, s0));\n\n    s += 1 * 4;\n    d += 1 * 4;\n    n -= 1;\n  }\n\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__src(\n    uint8_t* dst_ptr,\n    size_t dst_len,\n    uint8_t* dst_palette_ptr,\n    size_t dst_palette_len,\n    const uint8" +
 	"_t* src_ptr,\n    size_t src_len) {\n  size_t dst_len4 = dst_len / 4;\n  size_t src_len8 = src_len / 8;\n  size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n\n  size_t n = len;\n  while (n >= 1) {\n    wuffs_base__poke_u32le__no_bounds_check(\n        d + (0 * 4), wuffs_base__color_u64__as__color_u32(\n                         wuffs_base__peek_u64le__no_bounds_check(s + (0 * 8))));\n\n    s += 1 * 8;\n    d += 1 * 4;\n    n -= 1;\n  }\n  return len;\n}\n\nstatic uint64_t  //\nwuffs_base__pixel_swizzler__bgra_nonpremul__bgra_nonpremul_4x16le__src_over(\n    uint8_t* dst_ptr,\n    size_t dst_len,\n    uint8_t* dst_palette_ptr,\n    size_t dst_palette_len,\n    const uint8_t* src_ptr,\n    size_t src_len) {\n  size_t dst_len4 = dst_len / 4;\n  size_t src_len8 = src_len / 8;\n  size_t len = (dst_len4 < src_len8) ? dst_len4 : src_len8;\n  uint8_t* d = dst_ptr;\n  const uint8_t* s = src_ptr;\n  size_t n = len;\n\n  while (n >= 1) {\n    uint64_t d0 = wuffs_base__color_u32__as__color_u64(" +
diff --git a/release/c/wuffs-unsupported-snapshot.c b/release/c/wuffs-unsupported-snapshot.c
index b745f4f..1fc441d 100644
--- a/release/c/wuffs-unsupported-snapshot.c
+++ b/release/c/wuffs-unsupported-snapshot.c
@@ -16628,7 +16628,6 @@
     d[0] = s0;
     d[1] = s1;
     d[2] = s2;
-    d[3] = 0xFF;
 
     s += 1 * 4;
     d += 1 * 3;
@@ -16782,7 +16781,6 @@
     d[0] = s2;
     d[1] = s1;
     d[2] = s0;
-    d[3] = 0xFF;
 
     s += 1 * 4;
     d += 1 * 3;