Fix part of #34640: colors darkening when using the vertex paint blur tool.

The problem was that vertex colors only have 8 bits of precision, and integer
division always rounds down, so after some color blending iterations everything
gets darker. Instead use integer division that behaves like round() instead of
floor() for blending operations.

4 files changed, 62 insertions, 49 deletions

diff --git a/source/blender/blenlib/BLI_math_base.h b/source/blender/blenlib/BLI_math_base.h
index 67c1ffcebc0..04cf0b9d275 100644
--- a/source/blender/blenlib/BLI_math_base.h
+++ b/source/blender/blenlib/BLI_math_base.h
@@ -213,6 +213,8 @@ MINLINE int is_power_of_2_i(int n);
 MINLINE int power_of_2_max_i(int n);
 MINLINE int power_of_2_min_i(int n);
 
+MINLINE int divide_round_i(int a, int b);
+
 MINLINE float shell_angle_to_dist(const float angle);
 
 #if (defined(WIN32) || defined(WIN64)) && !defined(FREE_WINDOWS)
diff --git a/source/blender/blenlib/intern/math_base_inline.c b/source/blender/blenlib/intern/math_base_inline.c
index d7c950f651a..61af5b5079a 100644
--- a/source/blender/blenlib/intern/math_base_inline.c
+++ b/source/blender/blenlib/intern/math_base_inline.c
@@ -145,6 +145,13 @@ MINLINE int power_of_2_min_i(int n)
 	return n;
 }
 
+/* integer division that rounds 0.5 up, particularly useful for color blending
+ * with integers, to avoid gradual darkening when rounding down */
+MINLINE int divide_round_i(int a, int b)
+{
+	return (2*a + b)/(2*b);
+}
+
 MINLINE unsigned int highest_order_bit_i(unsigned int n)
 {
 	n |= (n >>  1);
diff --git a/source/blender/editors/sculpt_paint/paint_vertex.c b/source/blender/editors/sculpt_paint/paint_vertex.c
index d0f1bb882a3..7cc224db085 100644
--- a/source/blender/editors/sculpt_paint/paint_vertex.c
+++ b/source/blender/editors/sculpt_paint/paint_vertex.c
@@ -256,9 +256,9 @@ static void do_shared_vertex_tesscol(Mesh *me)
 	scol = scolmain;
 	while (a--) {
 		if (scol[0] > 1) {
-			scol[1] /= scol[0];
-			scol[2] /= scol[0];
-			scol[3] /= scol[0];
+			scol[1] = divide_round_i(scol[1], scol[0]);
+			scol[2] = divide_round_i(scol[2], scol[0]);
+			scol[3] = divide_round_i(scol[3], scol[0]);
 		}
 		scol += 4;
 	}
@@ -287,7 +287,7 @@ static void do_shared_vertexcol(Mesh *me, int do_tessface)
 {
 	const int use_face_sel = (me->editflag & ME_EDIT_PAINT_FACE_SEL);
 	MPoly *mp;
-	float (*scol)[4];
+	int (*scol)[4];
 	int i, j, has_shared = 0;
 
 	/* if no mloopcol: do not do */
@@ -295,7 +295,7 @@ static void do_shared_vertexcol(Mesh *me, int do_tessface)
 
 	if (me->mloopcol == NULL || me->totvert == 0 || me->totpoly == 0) return;
 
-	scol = MEM_callocN(sizeof(float) * me->totvert * 5, "scol");
+	scol = MEM_callocN(sizeof(int) * me->totvert * 5, "scol");
 
 	for (i = 0, mp = me->mpoly; i < me->totpoly; i++, mp++) {
 		if ((use_face_sel == FALSE) || (mp->flag & ME_FACE_SEL)) {
@@ -305,7 +305,7 @@ static void do_shared_vertexcol(Mesh *me, int do_tessface)
 				scol[ml->v][0] += lcol->r;
 				scol[ml->v][1] += lcol->g;
 				scol[ml->v][2] += lcol->b;
-				scol[ml->v][3] += 1.0f;
+				scol[ml->v][3] += 1;
 				has_shared = 1;
 			}
 		}
@@ -313,8 +313,10 @@ static void do_shared_vertexcol(Mesh *me, int do_tessface)
 
 	if (has_shared) {
 		for (i = 0; i < me->totvert; i++) {
-			if (scol[i][3] != 0.0f) {
-				mul_v3_fl(scol[i], 1.0f / scol[i][3]);
+			if (scol[i][3] != 0) {
+				scol[i][0] = divide_round_i(scol[i][0], scol[i][3]);
+				scol[i][1] = divide_round_i(scol[i][1], scol[i][3]);
+				scol[i][2] = divide_round_i(scol[i][2], scol[i][3]);
 			}
 		}
 
@@ -613,9 +615,9 @@ BLI_INLINE unsigned int mcol_blend(unsigned int col1, unsigned int col2, int fac
 	cp2 = (unsigned char *)&col2;
 	cp  = (unsigned char *)&col;
 
-	cp[0] = (mfac * cp1[0] + fac * cp2[0]) / 255;
-	cp[1] = (mfac * cp1[1] + fac * cp2[1]) / 255;
-	cp[2] = (mfac * cp1[2] + fac * cp2[2]) / 255;
+	cp[0] = divide_round_i((mfac * cp1[0] + fac * cp2[0]), 255);
+	cp[1] = divide_round_i((mfac * cp1[1] + fac * cp2[1]), 255);
+	cp[2] = divide_round_i((mfac * cp1[2] + fac * cp2[2]), 255);
 	cp[3] = 255;
 
 	return col;
@@ -635,11 +637,11 @@ BLI_INLINE unsigned int mcol_add(unsigned int col1, unsigned int col2, int fac)
 	cp2 = (unsigned char *)&col2;
 	cp  = (unsigned char *)&col;
 
-	temp = cp1[0] + ((fac * cp2[0]) / 255);
+	temp = cp1[0] + divide_round_i((fac * cp2[0]), 255);
 	cp[0] = (temp > 254) ? 255 : temp;
-	temp = cp1[1] + ((fac * cp2[1]) / 255);
+	temp = cp1[1] + divide_round_i((fac * cp2[1]), 255);
 	cp[1] = (temp > 254) ? 255 : temp;
-	temp = cp1[2] + ((fac * cp2[2]) / 255);
+	temp = cp1[2] + divide_round_i((fac * cp2[2]), 255);
 	cp[2] = (temp > 254) ? 255 : temp;
 	cp[3] = 255;
 	
@@ -660,11 +662,11 @@ BLI_INLINE unsigned int mcol_sub(unsigned int col1, unsigned int col2, int fac)
 	cp2 = (unsigned char *)&col2;
 	cp  = (unsigned char *)&col;
 
-	temp = cp1[0] - ((fac * cp2[0]) / 255);
+	temp = cp1[0] - divide_round_i((fac * cp2[0]), 255);
 	cp[0] = (temp < 0) ? 0 : temp;
-	temp = cp1[1] - ((fac * cp2[1]) / 255);
+	temp = cp1[1] - divide_round_i((fac * cp2[1]), 255);
 	cp[1] = (temp < 0) ? 0 : temp;
-	temp = cp1[2] - ((fac * cp2[2]) / 255);
+	temp = cp1[2] - divide_round_i((fac * cp2[2]), 255);
 	cp[2] = (temp < 0) ? 0 : temp;
 	cp[3] = 255;
 
@@ -688,9 +690,9 @@ BLI_INLINE unsigned int mcol_mul(unsigned int col1, unsigned int col2, int fac)
 	cp  = (unsigned char *)&col;
 
 	/* first mul, then blend the fac */
-	cp[0] = (mfac * cp1[0] + fac * ((cp2[0] * cp1[0]) / 255)) / 255;
-	cp[1] = (mfac * cp1[1] + fac * ((cp2[1] * cp1[1]) / 255)) / 255;
-	cp[2] = (mfac * cp1[2] + fac * ((cp2[2] * cp1[2]) / 255)) / 255;
+	cp[0] = divide_round_i(mfac * cp1[0] * 255 + fac * cp2[0] * cp1[0], 255*255);
+	cp[1] = divide_round_i(mfac * cp1[1] * 255 + fac * cp2[1] * cp1[1], 255*255);
+	cp[2] = divide_round_i(mfac * cp1[2] * 255 + fac * cp2[2] * cp1[2], 255*255);
 	cp[3] = 255;
 
 	return col;
@@ -721,9 +723,9 @@ BLI_INLINE unsigned int mcol_lighten(unsigned int col1, unsigned int col2, int f
 		return col1;
 	}
 
-	cp[0] = (mfac * cp1[0] + fac * cp2[0]) / 255;
-	cp[1] = (mfac * cp1[1] + fac * cp2[1]) / 255;
-	cp[2] = (mfac * cp1[2] + fac * cp2[2]) / 255;
+	cp[0] = divide_round_i(mfac * cp1[0] + fac * cp2[0], 255);
+	cp[1] = divide_round_i(mfac * cp1[1] + fac * cp2[1], 255);
+	cp[2] = divide_round_i(mfac * cp1[2] + fac * cp2[2], 255);
 	cp[3] = 255;
 
 	return col;
@@ -754,9 +756,9 @@ BLI_INLINE unsigned int mcol_darken(unsigned int col1, unsigned int col2, int fa
 		return col1;
 	}
 
-	cp[0] = (mfac * cp1[0] + fac * cp2[0]) / 255;
-	cp[1] = (mfac * cp1[1] + fac * cp2[1]) / 255;
-	cp[2] = (mfac * cp1[2] + fac * cp2[2]) / 255;
+	cp[0] = divide_round_i((mfac * cp1[0] + fac * cp2[0]), 255);
+	cp[1] = divide_round_i((mfac * cp1[1] + fac * cp2[1]), 255);
+	cp[2] = divide_round_i((mfac * cp1[2] + fac * cp2[2]), 255);
 	cp[3] = 255;
 	return col;
 }
@@ -2790,6 +2792,7 @@ static void vpaint_paint_poly(VPaint *vp, VPaintData *vpd, Mesh *me,
 	unsigned int *lcolorig = ((unsigned int *)vp->vpaint_prev) + mpoly->loopstart;
 	float alpha;
 	int i, j;
+	int totloop = mpoly->totloop;
 
 	int brush_alpha_pressure_i = (int)(brush_alpha_pressure * 255.0f);
 
@@ -2798,7 +2801,7 @@ static void vpaint_paint_poly(VPaint *vp, VPaintData *vpd, Mesh *me,
 		unsigned int tcol;
 		char *col;
 
-		for (j = 0; j < mpoly->totloop; j++) {
+		for (j = 0; j < totloop; j++) {
 			col = (char *)(lcol + j);
 			blend[0] += col[0];
 			blend[1] += col[1];
@@ -2806,10 +2809,10 @@ static void vpaint_paint_poly(VPaint *vp, VPaintData *vpd, Mesh *me,
 			blend[3] += col[3];
 		}
 
-		blend[0] /= mpoly->totloop;
-		blend[1] /= mpoly->totloop;
-		blend[2] /= mpoly->totloop;
-		blend[3] /= mpoly->totloop;
+		blend[0] = divide_round_i(blend[0], totloop);
+		blend[1] = divide_round_i(blend[1], totloop);
+		blend[2] = divide_round_i(blend[2], totloop);
+		blend[3] = divide_round_i(blend[3], totloop);
 		col = (char *)&tcol;
 		col[0] = blend[0];
 		col[1] = blend[1];
@@ -2820,7 +2823,7 @@ static void vpaint_paint_poly(VPaint *vp, VPaintData *vpd, Mesh *me,
 	}
 
 	ml = me->mloop + mpoly->loopstart;
-	for (i = 0; i < mpoly->totloop; i++, ml++) {
+	for (i = 0; i < totloop; i++, ml++) {
 		float rgba[4];
 		unsigned int paintcol;
 		alpha = calc_vp_alpha_col_dl(vp, vc, vpd->vpimat,
@@ -2851,7 +2854,7 @@ static void vpaint_paint_poly(VPaint *vp, VPaintData *vpd, Mesh *me,
 
 			ml = me->mloop + mpoly->loopstart;
 			mlc = me->mloopcol + mpoly->loopstart;
-			for (j = 0; j < mpoly->totloop; j++, ml++, mlc++) {
+			for (j = 0; j < totloop; j++, ml++, mlc++) {
 				if (ml->v == mf->v1) {
 					MESH_MLOOPCOL_TO_MCOL(mlc, mc + 0);
 				}
diff --git a/source/blender/imbuf/intern/rectop.c b/source/blender/imbuf/intern/rectop.c
index 79bf9eba1cc..d9375e3919b 100644
--- a/source/blender/imbuf/intern/rectop.c
+++ b/source/blender/imbuf/intern/rectop.c
@@ -34,6 +34,7 @@
 #include <stdlib.h>
 
 #include "BLI_utildefines.h"
+#include "BLI_math_base.h"
 #include "BLI_math_color.h"
 #include "BLI_math_vector.h"
 
@@ -54,20 +55,20 @@ static void blend_color_mix(char cp[3], const char cp1[3], const char cp2[3], co
 	 * errors that can turn colors black fast after repeated blending */
 	const int mfac = 255 - fac;
 
-	cp[0] = (mfac * cp1[0] + fac * cp2[0]) / 255;
-	cp[1] = (mfac * cp1[1] + fac * cp2[1]) / 255;
-	cp[2] = (mfac * cp1[2] + fac * cp2[2]) / 255;
+	cp[0] = divide_round_i((mfac * cp1[0] + fac * cp2[0]), 255);
+	cp[1] = divide_round_i((mfac * cp1[1] + fac * cp2[1]), 255);
+	cp[2] = divide_round_i((mfac * cp1[2] + fac * cp2[2]), 255);
 }
 
 static void blend_color_add(char cp[3], const char cp1[3], const char cp2[3], const int fac)
 {
 	int temp;
 
-	temp = cp1[0] + ((fac * cp2[0]) / 255);
+	temp = cp1[0] + divide_round_i(fac * cp2[0], 255);
 	if (temp > 254) cp[0] = 255; else cp[0] = temp;
-	temp = cp1[1] + ((fac * cp2[1]) / 255);
+	temp = cp1[1] + divide_round_i(fac * cp2[1] , 255);
 	if (temp > 254) cp[1] = 255; else cp[1] = temp;
-	temp = cp1[2] + ((fac * cp2[2]) / 255);
+	temp = cp1[2] + divide_round_i(fac * cp2[2] , 255);
 	if (temp > 254) cp[2] = 255; else cp[2] = temp;
 }
 
@@ -75,11 +76,11 @@ static void blend_color_sub(char cp[3], const char cp1[3], const char cp2[3], co
 {
 	int temp;
 
-	temp = cp1[0] - ((fac * cp2[0]) / 255);
+	temp = cp1[0] - divide_round_i(fac * cp2[0], 255);
 	if (temp < 0) cp[0] = 0; else cp[0] = temp;
-	temp = cp1[1] - ((fac * cp2[1]) / 255);
+	temp = cp1[1] - divide_round_i(fac * cp2[1], 255);
 	if (temp < 0) cp[1] = 0; else cp[1] = temp;
-	temp = cp1[2] - ((fac * cp2[2]) / 255);
+	temp = cp1[2] - divide_round_i(fac * cp2[2], 255);
 	if (temp < 0) cp[2] = 0; else cp[2] = temp;
 }
 
@@ -88,9 +89,9 @@ static void blend_color_mul(char cp[3], const char cp1[3], const char cp2[3], co
 	int mfac = 255 - fac;
 	
 	/* first mul, then blend the fac */
-	cp[0] = (mfac * cp1[0] + fac * ((cp1[0] * cp2[0]) / 255)) / 255;
-	cp[1] = (mfac * cp1[1] + fac * ((cp1[1] * cp2[1]) / 255)) / 255;
-	cp[2] = (mfac * cp1[2] + fac * ((cp1[2] * cp2[2]) / 255)) / 255;
+	cp[0] = divide_round_i((mfac * cp1[0] * 255) + (fac * cp1[0] * cp2[0]), 255*255);
+	cp[1] = divide_round_i((mfac * cp1[1] * 255) + (fac * cp1[1] * cp2[1]), 255*255);
+	cp[2] = divide_round_i((mfac * cp1[2] * 255) + (fac * cp1[2] * cp2[2]), 255*255);
 }
 
 static void blend_color_lighten(char cp[3], const char cp1[3], const char cp2[3], const int fac)
@@ -123,7 +124,7 @@ static void blend_color_darken(char cp[3], const char cp1[3], const char cp2[3],
 
 static void blend_color_erase_alpha(char cp[4], const char cp1[4], const char cp2[4], const int fac)
 {
-	int temp = (cp1[3] - fac * cp2[3] / 255);
+	int temp = divide_round_i(cp1[3] - fac * cp2[3], 255);
 
 	cp[0] = cp1[0];
 	cp[1] = cp1[1];
@@ -133,7 +134,7 @@ static void blend_color_erase_alpha(char cp[4], const char cp1[4], const char cp
 
 static void blend_color_add_alpha(char cp[4], const char cp1[4], const char cp2[4], const int fac)
 {
-	int temp = (cp1[3] + fac * cp2[3] / 255);
+	int temp = divide_round_i(cp1[3] + fac * cp2[3], 255);
 
 	cp[0] = cp1[0];
 	cp[1] = cp1[1];
@@ -175,11 +176,11 @@ unsigned int IMB_blend_color(unsigned int src1, unsigned int src2, int fac, IMB_
 	}
 
 	if (mode == IMB_BLEND_ERASE_ALPHA) {
-		temp = (cp1[3] - fac * cp2[3] / 255);
+		temp = divide_round_i(cp1[3] - fac * cp2[3], 255);
 		cp[3] = (temp < 0) ? 0 : temp;
 	}
 	else { /* this does ADD_ALPHA also */
-		temp = (cp1[3] + fac * cp2[3] / 255);
+		temp = divide_round_i(cp1[3] + fac * cp2[3], 255);
 		cp[3] = (temp > 255) ? 255 : temp;
 	}