--- rxvt-unicode/src/rxvtimg.C	2012/06/07 19:24:13	1.49
+++ rxvt-unicode/src/rxvtimg.C	2012/06/12 19:00:57	1.81
@@ -16,13 +16,6 @@
   ++ref->cnt;
 }
 
-#if 0
-rxvt_img::rxvt_img (rxvt_screen *screen, XRenderPictFormat *format, int width, int height, Pixmap pixmap)
-: s(screen), x(0), y(0), w(width), h(height), format(format), repeat(RepeatNormal), shared(false), pm(pixmap)
-{
-}
-#endif
-
 rxvt_img *
 rxvt_img::new_from_root (rxvt_screen *s)
 {
@@ -58,6 +51,8 @@
   return img;
 }
 
+# if HAVE_PIXBUF
+
 rxvt_img *
 rxvt_img::new_from_pixbuf (rxvt_screen *s, GdkPixbuf *pb)
 {
@@ -71,7 +66,8 @@
 
   // since we require rgb24/argb32 formats from xrender we assume
   // that both 24 and 32 bpp MUST be supported by any screen that supports xrender
-  int depth = gdk_pixbuf_get_has_alpha (pb) ? 32 : 24;
+
+  int byte_order = ecb_big_endian () ? MSBFirst : LSBFirst;
 
   XImage xi;
 
@@ -79,16 +75,19 @@
   xi.height           = height;
   xi.xoffset          = 0;
   xi.format           = ZPixmap;
-  xi.byte_order       = MSBFirst; // maybe go for host byte order, because servers are usually local?
-  xi.bitmap_unit      = 32;
-  xi.bitmap_bit_order = MSBFirst;
+  xi.byte_order       = ImageByteOrder (dpy);
+  xi.bitmap_unit      = 0;         //XY only, unused
+  xi.bitmap_bit_order = 0;         //XY only, unused
   xi.bitmap_pad       = BitmapPad (dpy);
-  xi.depth            = depth;
+  xi.depth            = 32;
   xi.bytes_per_line   = 0;
-  xi.bits_per_pixel   = 32;
-  xi.red_mask         = 0x00ff0000;
-  xi.green_mask       = 0x0000ff00;
-  xi.blue_mask        = 0x000000ff;
+  xi.bits_per_pixel   = 32;         //Z only
+  xi.red_mask         = 0x00000000; //Z only, unused
+  xi.green_mask       = 0x00000000; //Z only, unused
+  xi.blue_mask        = 0x00000000; //Z only, unused
+  xi.obdata           = 0;          // probably unused
+
+  bool byte_order_mismatch = byte_order != xi.byte_order;
 
   if (!XInitImage (&xi))
     rxvt_fatal ("unable to initialise ximage, please report.\n");
@@ -99,39 +98,51 @@
   xi.data = (char *)rxvt_malloc (height * xi.bytes_per_line);
 
   int rowstride = gdk_pixbuf_get_rowstride (pb);
-
-  assert (3 + (depth == 32) == gdk_pixbuf_get_n_channels (pb));
+  bool pb_has_alpha = gdk_pixbuf_get_has_alpha (pb);
   unsigned char *row = gdk_pixbuf_get_pixels (pb);
+
   char *line = xi.data;
 
   for (int y = 0; y < height; y++)
     {
-      unsigned char r, g, b, a;
-      unsigned char *data = row;
+      unsigned char *src = row;
+      uint32_t      *dst = (uint32_t *)line;
 
-      if (depth == 24)
+      if (!pb_has_alpha)
         for (int x = 0; x < width; x++)
           {
-            r = *data++;
-            g = *data++;
-            b = *data++;
-            *line++ = 0;
-            *line++ = r;
-            *line++ = g;
-            *line++ = b;
+            uint8_t r = *src++;
+            uint8_t g = *src++;
+            uint8_t b = *src++;
+
+            uint32_t v = (255 << 24) | (r << 16) | (g << 8) | b;
+            
+            if (ecb_big_endian () ? !byte_order_mismatch : byte_order_mismatch)
+              v = ecb_bswap32 (v);
+
+            *dst++ = v;
           }
       else
         for (int x = 0; x < width; x++)
           {
-            uint32_t v = *(uint32_t *)data; data += 4;
-            v = ecb_big_endian () ? ecb_rotr32 (v, 8) : ecb_rotl32 (v, 8);
-            *(uint32_t *)line = x; line += 4;
+            uint32_t v = *(uint32_t *)src; src += 4;
+
+            if (ecb_big_endian ())
+              v = ecb_bswap32 (v);
+
+            v = ecb_rotl32 (v, 8); // abgr to bgra
+
+            if (!byte_order_mismatch)
+              v = ecb_bswap32 (v);
+
+            *dst++ = v;
           }
 
       row += rowstride;
+      line += xi.bytes_per_line;
     }
 
-  rxvt_img *img = new rxvt_img (s, XRenderFindStandardFormat (dpy, depth == 24 ? PictStandardRGB24 : PictStandardARGB32), 0, 0, width, height);
+  rxvt_img *img = new rxvt_img (s, XRenderFindStandardFormat (dpy, PictStandardARGB32), 0, 0, width, height);
   img->alloc ();
 
   GC gc = XCreateGC (dpy, img->pm, 0, 0);
@@ -159,6 +170,8 @@
   return img;
 }
 
+# endif
+
 void
 rxvt_img::destroy ()
 {
@@ -201,7 +214,6 @@
   if (ref->cnt == 1 && ref->ours)
     return;
 
-  //TODO: maybe should reify instead
   Pixmap pm2 = XCreatePixmap (s->display->dpy, s->display->root, ref->w, ref->h, format->depth);
   GC gc = XCreateGC (s->display->dpy, pm, 0, 0);
   XCopyArea (s->display->dpy, pm, pm2, gc, 0, 0, ref->w, ref->h, 0, 0);
@@ -216,11 +228,14 @@
 void
 rxvt_img::fill (const rxvt_color &c)
 {
-  XGCValues gcv;
-  gcv.foreground = c;
-  GC gc = XCreateGC (s->display->dpy, pm, GCForeground, &gcv);
-  XFillRectangle (s->display->dpy, pm, gc, 0, 0, w, h);
-  XFreeGC (s->display->dpy, gc);
+  rgba cc;
+  c.get (cc);
+  XRenderColor rc = { cc.r, cc.g, cc.b, cc.a };
+
+  Display *dpy = s->display->dpy;
+  Picture src = src_picture ();
+  XRenderFillRectangle (dpy, PictOpSrc, src, &rc, 0, 0, w, h);
+  XRenderFreePicture (dpy, src);
 }
 
 static void
@@ -257,11 +272,10 @@
   rxvt_img *img = new rxvt_img (s, format, x, y, w, h, repeat);
   img->alloc ();
 
-  Picture src = src_picture ();
-
   XRenderPictureAttributes pa;
   pa.repeat = RepeatPad;
-  Picture dst = XRenderCreatePicture (dpy, img->pm, format, CPRepeat, &pa);
+  Picture src = XRenderCreatePicture (dpy, pm, format, CPRepeat, &pa);
+  Picture dst = XRenderCreatePicture (dpy, img->pm, format, 0, 0);
 
   Pixmap tmp_pm = XCreatePixmap (dpy, pm, w, h, format->depth);
   Picture tmp = XRenderCreatePicture (dpy, tmp_pm , format, CPRepeat, &pa);
@@ -287,7 +301,7 @@
       get_gaussian_kernel (rv, size, kernel, params);
       ::swap (params[0], params[1]);
 
-      XRenderSetPictureFilter (dpy, src, FilterConvolution, params, size+2);
+      XRenderSetPictureFilter (dpy, tmp, FilterConvolution, params, size+2);
       XRenderComposite (dpy,
                         PictOpSrc,
                         tmp,
@@ -301,6 +315,7 @@
 
   free (kernel);
   free (params);
+
   XRenderFreePicture (dpy, src);
   XRenderFreePicture (dpy, dst);
   XRenderFreePicture (dpy, tmp);
@@ -309,61 +324,120 @@
 }
 
 static Picture
-create_xrender_mask (Display *dpy, Drawable drawable, Bool argb)
+create_xrender_mask (Display *dpy, Drawable drawable, Bool argb, Bool component_alpha)
 {
   Pixmap pixmap = XCreatePixmap (dpy, drawable, 1, 1, argb ? 32 : 8);
 
   XRenderPictFormat *format = XRenderFindStandardFormat (dpy, argb ? PictStandardARGB32 : PictStandardA8);
   XRenderPictureAttributes pa;
-  pa.repeat = True;
-  Picture mask = XRenderCreatePicture (dpy, pixmap, format, CPRepeat, &pa);
+  pa.repeat = RepeatNormal;
+  pa.component_alpha = component_alpha;
+  Picture mask = XRenderCreatePicture (dpy, pixmap, format, CPRepeat | CPComponentAlpha, &pa);
 
   XFreePixmap (dpy, pixmap);
 
   return mask;
 }
 
+static void
+extract (int32_t cl0, int32_t cl1, int32_t &c, unsigned short &xc)
+{
+  int32_t x = clamp (c, cl0, cl1);
+  c -= x;
+  xc = x;
+}
+
+static bool
+extract (int32_t cl0, int32_t cl1, int32_t &r, int32_t &g, int32_t &b, int32_t &a, unsigned short &xr, unsigned short &xg, unsigned short &xb, unsigned short &xa)
+{
+  extract (cl0, cl1, r, xr);
+  extract (cl0, cl1, g, xg);
+  extract (cl0, cl1, b, xb);
+  extract (cl0, cl1, a, xa);
+
+  return xr | xg | xb | xa;
+}
+
 void
-rxvt_img::brightness (unsigned short r, unsigned short g, unsigned short b, unsigned short a)
+rxvt_img::brightness (int32_t r, int32_t g, int32_t b, int32_t a)
 {
+  unshare ();
+
   Display *dpy = s->display->dpy;
-  Picture src = create_xrender_mask (dpy, pm, True);
   Picture dst = XRenderCreatePicture (dpy, pm, format, 0, 0);
 
-  XRenderColor mask_c;
-  mask_c.red   = r;
-  mask_c.green = g;
-  mask_c.blue  = b;
-  mask_c.alpha = a;
-  XRenderFillRectangle (dpy, PictOpSrc, src, &mask_c, 0, 0, 1, 1);
+  // loop should not be needed for brightness, as only -1..1 makes sense
+  //while (r | g | b | a)
+    {
+      unsigned short xr, xg, xb, xa;
+      XRenderColor mask_c;
 
-  XRenderComposite (dpy, PictOpAdd, src, None, dst, 0, 0, 0, 0, 0, 0, w, h);
+      if (extract (0, 65535, r, g, b, a, mask_c.red, mask_c.green, mask_c.blue, mask_c.alpha))
+        XRenderFillRectangle (dpy, PictOpAdd, dst, &mask_c, 0, 0, w, h);
+
+      if (extract (-65535, 0, r, g, b, a, mask_c.red, mask_c.green, mask_c.blue, mask_c.alpha))
+        {
+          XRenderColor mask_w = { 65535, 65535, 65535, 65535 };
+          XRenderFillRectangle (dpy, PictOpDifference, dst, &mask_w, 0, 0, w, h);
+          mask_c.red   = -mask_c.red; //TODO: verify that doing clamp, assign, and negation does the right thing
+          mask_c.green = -mask_c.green;
+          mask_c.blue  = -mask_c.blue;
+          mask_c.alpha = -mask_c.alpha;
+          XRenderFillRectangle (dpy, PictOpAdd, dst, &mask_c, 0, 0, w, h);
+          XRenderFillRectangle (dpy, PictOpDifference, dst, &mask_w, 0, 0, w, h);
+        }
+    }
 
-  XRenderFreePicture (dpy, src);
   XRenderFreePicture (dpy, dst);
 }
 
 void
-rxvt_img::contrast (unsigned short r, unsigned short g, unsigned short b, unsigned short a)
+rxvt_img::contrast (int32_t r, int32_t g, int32_t b, int32_t a)
 {
-  if (!(s->display->flags & DISPLAY_HAS_RENDER_MUL))
-    return;
+  if (r < 0 || g < 0 || b < 0 || a < 0)
+    rxvt_fatal ("rxvt_img::contrast does not support negative values.\n");
+
+  rxvt_img *img = new rxvt_img (s, format, x, y, w, h, repeat);
+  img->alloc ();
+
+  {
+    rxvt_color empty;
+    empty.set (s, rgba (0, 0, 0, 0));
+    img->fill (empty);
+  }
+
+  // premultiply (yeah, these are not exact, sue me or fix it)
+  r = (r * (a >> 8)) >> 8;
+  g = (g * (a >> 8)) >> 8;
+  b = (b * (a >> 8)) >> 8;
 
   Display *dpy = s->display->dpy;
-  Picture src = create_xrender_mask (dpy, pm, True);
-  Picture dst = XRenderCreatePicture (dpy, pm, format, 0, 0);
 
-  XRenderColor mask_c;
-  mask_c.red   = r;
-  mask_c.green = g;
-  mask_c.blue  = b;
-  mask_c.alpha = a;
-  XRenderFillRectangle (dpy, PictOpSrc, src, &mask_c, 0, 0, 1, 1);
+  Picture src = src_picture ();
+  Picture dst = XRenderCreatePicture (dpy, img->pm, format, 0, 0);
+  Picture mul = create_xrender_mask (dpy, pm, True, True);
 
-  XRenderComposite (dpy, PictOpMultiply, src, None, dst, 0, 0, 0, 0, 0, 0, w, h);
+  //TODO: this operator does not yet implement some useful contrast
+  while (r | g | b | a)
+    {
+      unsigned short xr, xg, xb, xa;
+      XRenderColor mask_c;
 
-  XRenderFreePicture (dpy, src);
+      if (extract (0, 65535, r, g, b, a, mask_c.red, mask_c.green, mask_c.blue, mask_c.alpha))
+        {
+          XRenderFillRectangle (dpy, PictOpSrc, mul, &mask_c, 0, 0, 1, 1);
+          XRenderComposite (dpy, PictOpAdd, src, mul, dst, 0, 0, 0, 0, 0, 0, w, h);
+        }
+    }
+
+  XRenderFreePicture (dpy, mul);
   XRenderFreePicture (dpy, dst);
+  XRenderFreePicture (dpy, src);
+
+  ::swap (img->ref, ref);
+  ::swap (img->pm , pm );
+
+  delete img;
 }
 
 rxvt_img *
@@ -400,9 +474,10 @@
 
   Display *dpy = s->display->dpy;
 
-  bool alpha = !format->direct.alphaMask
-               && (x || y)
-               && repeat == RepeatNone;
+  // add an alpha channel if...
+  bool alpha = !format->direct.alphaMask // pixmap has none yet
+               && (x || y)               // we need one because of non-zero offset
+               && repeat == RepeatNone;  // and we have no good pixels to fill with
 
   rxvt_img *img = new rxvt_img (s, alpha ? find_alpha_format_for (dpy, format) : format, 0, 0, w, h, repeat);
   img->alloc ();
@@ -446,10 +521,66 @@
   return img;
 }
 
+static void
+mat_invert (double mat[3][3], double (&inv)[3][3])
+{
+  double s0 = mat [2][2] * mat [1][1] - mat [2][1] * mat [1][2];
+  double s1 = mat [2][1] * mat [0][2] - mat [2][2] * mat [0][1];
+  double s2 = mat [1][2] * mat [0][1] - mat [1][1] * mat [0][2];
+
+  double invdet = 1. / (mat [0][0] * s0 + mat [1][0] * s1 + mat [2][0] * s2);
+
+  inv [0][0] = invdet * s0;
+  inv [0][1] = invdet * s1;
+  inv [0][2] = invdet * s2;
+
+  inv [1][0] = invdet * (mat [2][0] * mat [1][2] - mat [2][2] * mat [1][0]);
+  inv [1][1] = invdet * (mat [2][2] * mat [0][0] - mat [2][0] * mat [0][2]);
+  inv [1][2] = invdet * (mat [1][0] * mat [0][2] - mat [1][2] * mat [0][0]);
+
+  inv [2][0] = invdet * (mat [2][1] * mat [1][0] - mat [2][0] * mat [1][1]);
+  inv [2][1] = invdet * (mat [2][0] * mat [0][1] - mat [2][1] * mat [0][0]);
+  inv [2][2] = invdet * (mat [1][1] * mat [0][0] - mat [1][0] * mat [0][1]);
+}
+
+static double
+mat_apply (double mat[3][3], int i, double x, double y)
+{
+  double v = mat [i][0] * x + mat [i][1] * y + mat [i][2];
+  double w = mat [2][0] * x + mat [2][1] * y + mat [2][2];
+
+  return v * (1. / w);
+}
+
 rxvt_img *
-rxvt_img::transform (int new_width, int new_height, double matrix[9])
+rxvt_img::transform (double matrix[3][3])
 {
-  rxvt_img *img = new rxvt_img (s, format, 0, 0, new_width, new_height, repeat);
+  // find new offset
+  int ox = mat_apply (matrix, 0, x, y);
+  int oy = mat_apply (matrix, 1, x, y);
+
+  // calculate new pixel bounding box coordinates
+  double d [2], rmin[2], rmax[2];
+
+  for (int i = 0; i < 2; ++i)
+    {
+      double v;
+      v = mat_apply (matrix, i, 0, 0);         rmin [i] =            rmax [i] = v; d [i] = v;
+      v = mat_apply (matrix, i, w, 0); min_it (rmin [i], v); max_it (rmax [i],  v);
+      v = mat_apply (matrix, i, 0, h); min_it (rmin [i], v); max_it (rmax [i],  v);
+      v = mat_apply (matrix, i, w, h); min_it (rmin [i], v); max_it (rmax [i],  v);
+    }
+
+  int dx = floor (rmin [0]);
+  int dy = floor (rmin [1]);
+
+  int new_width  = ceil (rmax [0] - dx);
+  int new_height = ceil (rmax [1] - dy);
+
+  double inv[3][3];
+  mat_invert (matrix, inv);
+
+  rxvt_img *img = new rxvt_img (s, format, ox - dx - d [0], oy - dy - d [1], new_width, new_height, repeat);
   img->alloc ();
 
   Display *dpy = s->display->dpy;
@@ -460,16 +591,11 @@
 
   for (int i = 0; i < 3; ++i)
     for (int j = 0; j < 3; ++j)
-      xfrm.matrix [i][j] = XDoubleToFixed (matrix [i * 3 + j]);
-
-#if 0
-  xfrm.matrix [0][2] -= XDoubleToFixed (x);//TODO
-  xfrm.matrix [1][2] -= XDoubleToFixed (y);
-#endif
+      xfrm.matrix [i][j] = XDoubleToFixed (inv [i][j]);
 
   XRenderSetPictureFilter (dpy, src, "good", 0, 0);
   XRenderSetPictureTransform (dpy, src, &xfrm);
-  XRenderComposite (dpy, PictOpSrc, src, None, dst, 0, 0, 0, 0, 0, 0, new_width, new_height);
+  XRenderComposite (dpy, PictOpSrc, src, None, dst, dx, dy, 0, 0, 0, 0, new_width, new_height);
 
   XRenderFreePicture (dpy, src);
   XRenderFreePicture (dpy, dst);
@@ -483,16 +609,16 @@
   if (w == new_width && h == new_height)
     return clone ();
 
-  double matrix[9] = {
-    w  / (double)new_width, 0, 0,
-    0, h / (double)new_height, 0,
-    0,                      0, 1
+  double matrix[3][3] = {
+    { new_width / (double)w,  0, 0 },
+    { 0, new_height / (double)h, 0 },
+    { 0,                      0, 1 }
   };
 
   int old_repeat_mode = repeat;
-  repeat = RepeatPad; // not right, but xrender can't proeprly scale it seems
+  repeat = RepeatPad; // not right, but xrender can't properly scale it seems
 
-  rxvt_img *img = transform (new_width, new_height, matrix);
+  rxvt_img *img = transform (matrix);
 
   repeat = old_repeat_mode;
   img->repeat = repeat;
@@ -501,18 +627,26 @@
 }
 
 rxvt_img *
-rxvt_img::rotate (int new_width, int new_height, int x, int y, double phi)
+rxvt_img::rotate (int cx, int cy, double phi)
 {
   double s = sin (phi);
   double c = cos (phi);
 
-  double matrix[9] = {
-    c, -s, -c * x + s * y + x,
-    s,  c, -s * x - c * y + y,
-    0,  0,                  1
+  double matrix[3][3] = {
+    { c, -s, cx - c * cx + s * cy },
+    { s,  c, cy - s * cx - c * cy },
+    { 0,  0,                     1 }
+    //{ c, -s, 0 },
+    //{ s,  c, 0 },
+    //{ 0,  0, 1 }
   };
 
-  return transform (new_width, new_height, matrix);
+  //move (-cx, -cy);
+  rxvt_img *img = transform (matrix);
+  //move ( cx,  cy);
+  //img->move (cx, cy);
+
+  return img;
 }
 
 rxvt_img *
@@ -556,7 +690,7 @@
   Display *dpy = s->display->dpy;
   Picture src = img->src_picture ();
   Picture dst = XRenderCreatePicture (dpy, img2->pm, img2->format, 0, 0);
-  Picture mask = create_xrender_mask (dpy, img->pm, False);
+  Picture mask = create_xrender_mask (dpy, img->pm, False, False);
 
   XRenderColor mask_c;