swaybg: split into standalone project

The new upstream is https://github.com/swaywm/swaybg

This commit also refactors our use of gdk-pixbuf a bit, since the only
remaining reverse dependency is swaybar tray support.

1 files changed, 100 insertions, 0 deletions

diff --git a/common/background-image.c b/common/background-image.c
index 4431b725..de42e8e9 100644
--- a/common/background-image.c
+++ b/common/background-image.c
@@ -2,6 +2,9 @@
 #include "background-image.h"
 #include "cairo.h"
 #include "log.h"
+#if HAVE_GDK_PIXBUF
+#include <gdk-pixbuf/gdk-pixbuf.h>
+#endif
 
 enum background_mode parse_background_mode(const char *mode) {
         if (strcmp(mode, "stretch") == 0) {
@@ -21,6 +24,103 @@ enum background_mode parse_background_mode(const char *mode) {
         return BACKGROUND_MODE_INVALID;
 }
 
+#if HAVE_GDK_PIXBUF
+static cairo_surface_t* gdk_cairo_image_surface_create_from_pixbuf(
+                const GdkPixbuf *gdkbuf) {
+        int chan = gdk_pixbuf_get_n_channels(gdkbuf);
+        if (chan < 3) {
+                return NULL;
+        }
+
+        const guint8* gdkpix = gdk_pixbuf_read_pixels(gdkbuf);
+        if (!gdkpix) {
+                return NULL;
+        }
+        gint w = gdk_pixbuf_get_width(gdkbuf);
+        gint h = gdk_pixbuf_get_height(gdkbuf);
+        int stride = gdk_pixbuf_get_rowstride(gdkbuf);
+
+        cairo_format_t fmt = (chan == 3) ? CAIRO_FORMAT_RGB24 : CAIRO_FORMAT_ARGB32;
+        cairo_surface_t * cs = cairo_image_surface_create (fmt, w, h);
+        cairo_surface_flush (cs);
+        if ( !cs || cairo_surface_status(cs) != CAIRO_STATUS_SUCCESS) {
+                return NULL;
+        }
+
+        int cstride = cairo_image_surface_get_stride(cs);
+        unsigned char * cpix = cairo_image_surface_get_data(cs);
+
+        if (chan == 3) {
+                int i;
+                for (i = h; i; --i) {
+                        const guint8 *gp = gdkpix;
+                        unsigned char *cp = cpix;
+                        const guint8* end = gp + 3*w;
+                        while (gp < end) {
+#if G_BYTE_ORDER == G_LITTLE_ENDIAN
+                                cp[0] = gp[2];
+                                cp[1] = gp[1];
+                                cp[2] = gp[0];
+#else
+                                cp[1] = gp[0];
+                                cp[2] = gp[1];
+                                cp[3] = gp[2];
+#endif
+                                gp += 3;
+                                cp += 4;
+                        }
+                        gdkpix += stride;
+                        cpix += cstride;
+                }
+        } else {
+                /* premul-color = alpha/255 * color/255 * 255 = (alpha*color)/255
+                 * (z/255) = z/256 * 256/255     = z/256 (1 + 1/255)
+                 *         = z/256 + (z/256)/255 = (z + z/255)/256
+                 *         # recurse once
+                 *         = (z + (z + z/255)/256)/256
+                 *         = (z + z/256 + z/256/255) / 256
+                 *         # only use 16bit uint operations, loose some precision,
+                 *         # result is floored.
+                 *       ->  (z + z>>8)>>8
+                 *         # add 0x80/255 = 0.5 to convert floor to round
+                 *       =>  (z+0x80 + (z+0x80)>>8 ) >> 8
+                 * ------
+                 * tested as equal to lround(z/255.0) for uint z in [0..0xfe02]
+                 */
+#define PREMUL_ALPHA(x,a,b,z) \
+                G_STMT_START { z = a * b + 0x80; x = (z + (z >> 8)) >> 8; } \
+                G_STMT_END
+                int i;
+                for (i = h; i; --i) {
+                        const guint8 *gp = gdkpix;
+                        unsigned char *cp = cpix;
+                        const guint8* end = gp + 4*w;
+                        guint z1, z2, z3;
+                        while (gp < end) {
+#if G_BYTE_ORDER == G_LITTLE_ENDIAN
+                                PREMUL_ALPHA(cp[0], gp[2], gp[3], z1);
+                                PREMUL_ALPHA(cp[1], gp[1], gp[3], z2);
+                                PREMUL_ALPHA(cp[2], gp[0], gp[3], z3);
+                                cp[3] = gp[3];
+#else
+                                PREMUL_ALPHA(cp[1], gp[0], gp[3], z1);
+                                PREMUL_ALPHA(cp[2], gp[1], gp[3], z2);
+                                PREMUL_ALPHA(cp[3], gp[2], gp[3], z3);
+                                cp[0] = gp[3];
+#endif
+                                gp += 4;
+                                cp += 4;
+                        }
+                        gdkpix += stride;
+                        cpix += cstride;
+                }
+#undef PREMUL_ALPHA
+        }
+        cairo_surface_mark_dirty(cs);
+        return cs;
+}
+#endif // HAVE_GDK_PIXBUF
+
 cairo_surface_t *load_background_image(const char *path) {
         cairo_surface_t *image;
 #if HAVE_GDK_PIXBUF