2.50:

svn merge https://svn.blender.org/svnroot/bf-blender/trunk/blender -r19820:HEAD

Notes:
* Game and sequencer RNA, and sequencer header are now out of date
  a bit after changes in trunk.
* I didn't know how to port these bugfixes, most likely they are
  not needed anymore.
  * Fix "duplicate strip" always increase the user count for ipo.
  * IPO pinning on sequencer strips was lost during Undo.

6 files changed, 127 insertions, 48 deletions

diff --git a/source/blender/imbuf/IMB_imbuf.h b/source/blender/imbuf/IMB_imbuf.h
index 8bc1439fd09..1d8035a2358 100644
--- a/source/blender/imbuf/IMB_imbuf.h
+++ b/source/blender/imbuf/IMB_imbuf.h
@@ -410,6 +410,7 @@ void bilinear_interpolation(struct ImBuf *in, struct ImBuf *out, float u, float
 void bicubic_interpolation_color(struct ImBuf *in, unsigned char *col, float *col_float, float u, float v);
 void neareast_interpolation_color(struct ImBuf *in, unsigned char *col, float *col_float, float u, float v);
 void bilinear_interpolation_color(struct ImBuf *in, unsigned char *col, float *col_float, float u, float v);
+void bilinear_interpolation_color_wrap(struct ImBuf *in, unsigned char *col, float *col_float, float u, float v);
 
 /**
  * Change the ordering of the color bytes pointed to by rect from
diff --git a/source/blender/imbuf/intern/anim.c b/source/blender/imbuf/intern/anim.c
index 30f24d9bbf3..a2dbdfbe483 100644
--- a/source/blender/imbuf/intern/anim.c
+++ b/source/blender/imbuf/intern/anim.c
@@ -1046,10 +1046,11 @@ struct ImBuf * IMB_anim_absolute(struct anim * anim, int position) {
 	char head[256], tail[256];
 	unsigned short digits;
 	int pic;
-	int filter_y = (anim->ib_flags & IB_animdeinterlace);
-
+	int filter_y;
 	if (anim == NULL) return(0);
 
+	filter_y = (anim->ib_flags & IB_animdeinterlace);
+
 	if (anim->curtype == 0)	{
 		ibuf = anim_getnew(anim);
 		if (ibuf == NULL) {
diff --git a/source/blender/imbuf/intern/imageprocess.c b/source/blender/imbuf/intern/imageprocess.c
index fe7e26eac2b..e4977c77155 100644
--- a/source/blender/imbuf/intern/imageprocess.c
+++ b/source/blender/imbuf/intern/imageprocess.c
@@ -294,10 +294,79 @@ void bilinear_interpolation_color(struct ImBuf *in, unsigned char *outI, float *
 		b= v-floor(v);
 		a_b= a*b; ma_b= (1.0f-a)*b; a_mb= a*(1.0f-b); ma_mb= (1.0f-a)*(1.0f-b);
 		
-		outI[0]= ma_mb*row1I[0] + a_mb*row3I[0] + ma_b*row2I[0]+ a_b*row4I[0];
-		outI[1]= ma_mb*row1I[1] + a_mb*row3I[1] + ma_b*row2I[1]+ a_b*row4I[1];
-		outI[2]= ma_mb*row1I[2] + a_mb*row3I[2] + ma_b*row2I[2]+ a_b*row4I[2];
-		outI[3]= ma_mb*row1I[3] + a_mb*row3I[3] + ma_b*row2I[3]+ a_b*row4I[3];
+		/* need to add 0.5 to avoid rounding down (causes darken with the smear brush)
+		 * tested with white images and this should not wrap back to zero */
+		outI[0]= (ma_mb*row1I[0] + a_mb*row3I[0] + ma_b*row2I[0]+ a_b*row4I[0]) + 0.5f;
+		outI[1]= (ma_mb*row1I[1] + a_mb*row3I[1] + ma_b*row2I[1]+ a_b*row4I[1]) + 0.5f;
+		outI[2]= (ma_mb*row1I[2] + a_mb*row3I[2] + ma_b*row2I[2]+ a_b*row4I[2]) + 0.5f;
+		outI[3]= (ma_mb*row1I[3] + a_mb*row3I[3] + ma_b*row2I[3]+ a_b*row4I[3]) + 0.5f;
+	}
+}
+
+/* function assumes out to be zero'ed, only does RGBA */
+/* BILINEAR INTERPOLATION */
+
+/* Note about wrapping, the u/v still needs to be within the image bounds,
+ * just the interpolation is wrapped.
+ * This the same as bilinear_interpolation_color except it wraps rather then using empty and emptyI */
+void bilinear_interpolation_color_wrap(struct ImBuf *in, unsigned char *outI, float *outF, float u, float v)
+{
+	float *row1, *row2, *row3, *row4, a, b;
+	unsigned char *row1I, *row2I, *row3I, *row4I;
+	float a_b, ma_b, a_mb, ma_mb;
+	int y1, y2, x1, x2;
+	
+	
+	/* ImBuf in must have a valid rect or rect_float, assume this is alredy checked */
+
+	x1= (int)floor(u);
+	x2= (int)ceil(u);
+	y1= (int)floor(v);
+	y2= (int)ceil(v);
+
+	// sample area entirely outside image? 
+	if (x2<0 || x1>in->x-1 || y2<0 || y1>in->y-1) return;
+	
+	/* wrap interpolation pixels - main difference from bilinear_interpolation_color  */
+	if(x1<0)x1= in->x+x1;
+	if(y1<0)y1= in->y+y1;
+	
+	if(x2>=in->x)x2= x2-in->x;
+	if(y2>=in->y)y2= y2-in->y;
+
+	if (outF) {
+		// sample including outside of edges of image 
+		row1= (float *)in->rect_float + in->x * y1 * 4 + 4*x1;
+		row2= (float *)in->rect_float + in->x * y2 * 4 + 4*x1;
+		row3= (float *)in->rect_float + in->x * y1 * 4 + 4*x2;
+		row4= (float *)in->rect_float + in->x * y2 * 4 + 4*x2;
+		
+		a= u-floor(u);
+		b= v-floor(v);
+		a_b= a*b; ma_b= (1.0f-a)*b; a_mb= a*(1.0f-b); ma_mb= (1.0f-a)*(1.0f-b);
+		
+		outF[0]= ma_mb*row1[0] + a_mb*row3[0] + ma_b*row2[0]+ a_b*row4[0];
+		outF[1]= ma_mb*row1[1] + a_mb*row3[1] + ma_b*row2[1]+ a_b*row4[1];
+		outF[2]= ma_mb*row1[2] + a_mb*row3[2] + ma_b*row2[2]+ a_b*row4[2];
+		outF[3]= ma_mb*row1[3] + a_mb*row3[3] + ma_b*row2[3]+ a_b*row4[3];
+	}
+	if (outI) {
+		// sample including outside of edges of image 
+		row1I= (unsigned char *)in->rect + in->x * y1 * 4 + 4*x1;
+		row2I= (unsigned char *)in->rect + in->x * y2 * 4 + 4*x1;
+		row3I= (unsigned char *)in->rect + in->x * y1 * 4 + 4*x2;
+		row4I= (unsigned char *)in->rect + in->x * y2 * 4 + 4*x2;
+		
+		a= u-floor(u);
+		b= v-floor(v);
+		a_b= a*b; ma_b= (1.0f-a)*b; a_mb= a*(1.0f-b); ma_mb= (1.0f-a)*(1.0f-b);
+		
+		/* need to add 0.5 to avoid rounding down (causes darken with the smear brush)
+		 * tested with white images and this should not wrap back to zero */
+		outI[0]= (ma_mb*row1I[0] + a_mb*row3I[0] + ma_b*row2I[0]+ a_b*row4I[0]) + 0.5f;
+		outI[1]= (ma_mb*row1I[1] + a_mb*row3I[1] + ma_b*row2I[1]+ a_b*row4I[1]) + 0.5f;
+		outI[2]= (ma_mb*row1I[2] + a_mb*row3I[2] + ma_b*row2I[2]+ a_b*row4I[2]) + 0.5f;
+		outI[3]= (ma_mb*row1I[3] + a_mb*row3I[3] + ma_b*row2I[3]+ a_b*row4I[3]) + 0.5f;
 	}
 }
 
diff --git a/source/blender/imbuf/intern/openexr/openexr_api.cpp b/source/blender/imbuf/intern/openexr/openexr_api.cpp
index e723609f3ae..32d97d79bd7 100644
--- a/source/blender/imbuf/intern/openexr/openexr_api.cpp
+++ b/source/blender/imbuf/intern/openexr/openexr_api.cpp
@@ -570,13 +570,17 @@ void IMB_exr_write_channels(void *handle)
 	FrameBuffer frameBuffer;
 	ExrChannel *echan;
 	
-	for(echan= (ExrChannel *)data->channels.first; echan; echan= echan->next)
-		frameBuffer.insert (echan->name, Slice (FLOAT,  (char *)echan->rect, 
-												echan->xstride*sizeof(float), echan->ystride*sizeof(float)));
-	
-	data->ofile->setFrameBuffer (frameBuffer);
-	data->ofile->writePixels (data->height);	
-	
+	if(data->channels.first) {
+		for(echan= (ExrChannel *)data->channels.first; echan; echan= echan->next)
+			frameBuffer.insert (echan->name, Slice (FLOAT,  (char *)echan->rect, 
+													echan->xstride*sizeof(float), echan->ystride*sizeof(float)));
+		
+		data->ofile->setFrameBuffer (frameBuffer);
+		data->ofile->writePixels (data->height);	
+	}
+	else {
+		printf("Error: attempt to save MultiLayer without layers.\n");
+	}
 }
 
 void IMB_exr_read_channels(void *handle)
@@ -861,6 +865,7 @@ static const char *exr_rgba_channelname(InputFile *file, const char *chan)
 	
 	for (ChannelList::ConstIterator i = channels.begin(); i != channels.end(); ++i)
 	{
+		/* const Channel &channel = i.channel(); */ /* Not used yet */
 		const char *str= i.name();
 		int len= strlen(str);
 		if(len) {
diff --git a/source/blender/imbuf/intern/rotate.c b/source/blender/imbuf/intern/rotate.c
index c04987b3e71..732c06907df 100644
--- a/source/blender/imbuf/intern/rotate.c
+++ b/source/blender/imbuf/intern/rotate.c
@@ -30,6 +30,7 @@
  */
 
 #include "BLI_blenlib.h"
+#include  "BKE_utildefines.h"
 
 #include "imbuf.h"
 #include "imbuf_patch.h"
@@ -94,7 +95,6 @@ void IMB_flipy(struct ImBuf * ibuf)
 void IMB_flipx(struct ImBuf * ibuf)
 {
 	short x, y, xr, xl, yi;
-	unsigned int px;
 	float px_f[4];
 	
 	if (ibuf == NULL) return;
@@ -105,9 +105,7 @@ void IMB_flipx(struct ImBuf * ibuf)
 	if (ibuf->rect) {
 		for(yi=y-1;yi>=0;yi--) {
 			for(xr=x-1, xl=0; xr>=xl; xr--, xl++) {
-				px = ibuf->rect[(x*yi)+xr];
-				ibuf->rect[(x*yi)+xr] =	ibuf->rect[(x*yi)+xl];
-				ibuf->rect[(x*yi)+xl] =	px;		
+				SWAP(unsigned int, ibuf->rect[(x*yi)+xr], ibuf->rect[(x*yi)+xl]);
 			}
 		}
 	}
diff --git a/source/blender/imbuf/intern/scaling.c b/source/blender/imbuf/intern/scaling.c
index 807b0c84e90..1ccdad05deb 100644
--- a/source/blender/imbuf/intern/scaling.c
+++ b/source/blender/imbuf/intern/scaling.c
@@ -596,12 +596,12 @@ static void shrink_picture_byte(
 	y_counter = 65536;
 	for (y_src = 0; y_src < src_height; y_src++) {
 		unsigned char* line = src + y_src * 4 * src_width;
-		uintptr_t weight1y = 65536 - (y_dst & 0xffff);
-		uintptr_t weight2y = 65536 - weight1y;
+		uintptr_t weight1y = 65535 - (y_dst & 0xffff);
+		uintptr_t weight2y = 65535 - weight1y;
 		x_dst = 0;
 		for (x_src = 0; x_src < src_width; x_src++) {
-			uintptr_t weight1x = 65536 - (x_dst & 0xffff);
-			uintptr_t weight2x = 65536 - weight1x;
+			uintptr_t weight1x = 65535 - (x_dst & 0xffff);
+			uintptr_t weight2x = 65535 - weight1x;
 
 			uintptr_t x = x_dst >> 16;
 
@@ -609,34 +609,35 @@ static void shrink_picture_byte(
 
 			w = (weight1y * weight1x) >> 16;
 
-			dst_line1[x].r += (line[0] * w) >> 16;
-			dst_line1[x].g += (line[1] * w) >> 16;
-			dst_line1[x].b += (line[2] * w) >> 16;
-			dst_line1[x].a += (line[3] * w) >> 16;
+			/* ensure correct rounding, without this you get ugly banding, or too low color values (ton) */
+			dst_line1[x].r += (line[0] * w + 32767) >> 16;
+			dst_line1[x].g += (line[1] * w + 32767) >> 16;
+			dst_line1[x].b += (line[2] * w + 32767) >> 16;
+			dst_line1[x].a += (line[3] * w + 32767) >> 16;
 			dst_line1[x].weight += w;
 
 			w = (weight2y * weight1x) >> 16;
 
-			dst_line2[x].r += (line[0] * w) >> 16;
-			dst_line2[x].g += (line[1] * w) >> 16;
-			dst_line2[x].b += (line[2] * w) >> 16;
-			dst_line2[x].a += (line[3] * w) >> 16;
+			dst_line2[x].r += (line[0] * w + 32767) >> 16;
+			dst_line2[x].g += (line[1] * w + 32767) >> 16;
+			dst_line2[x].b += (line[2] * w + 32767) >> 16;
+			dst_line2[x].a += (line[3] * w + 32767) >> 16;
 			dst_line2[x].weight += w;
 
 			w = (weight1y * weight2x) >> 16;
 
-			dst_line1[x+1].r += (line[0] * w) >> 16;
-			dst_line1[x+1].g += (line[1] * w) >> 16;
-			dst_line1[x+1].b += (line[2] * w) >> 16;
-			dst_line1[x+1].a += (line[3] * w) >> 16;
+			dst_line1[x+1].r += (line[0] * w + 32767) >> 16;
+			dst_line1[x+1].g += (line[1] * w + 32767) >> 16;
+			dst_line1[x+1].b += (line[2] * w + 32767) >> 16;
+			dst_line1[x+1].a += (line[3] * w + 32767) >> 16;
 			dst_line1[x+1].weight += w;
 
 			w = (weight2y * weight2x) >> 16;
 
-			dst_line2[x+1].r += (line[0] * w) >> 16;
-			dst_line2[x+1].g += (line[1] * w) >> 16;
-			dst_line2[x+1].b += (line[2] * w) >> 16;
-			dst_line2[x+1].a += (line[3] * w) >> 16;
+			dst_line2[x+1].r += (line[0] * w + 32767) >> 16;
+			dst_line2[x+1].g += (line[1] * w + 32767) >> 16;
+			dst_line2[x+1].b += (line[2] * w + 32767) >> 16;
+			dst_line2[x+1].a += (line[3] * w + 32767) >> 16;
 			dst_line2[x+1].weight += w;
 
 			x_dst += dx_dst;
@@ -646,18 +647,18 @@ static void shrink_picture_byte(
 		y_dst += dy_dst;
 		y_counter -= dy_dst;
 		if (y_counter < 0) {
+			int val;
 			uintptr_t x;
 			struct scale_outpix_byte * temp;
 
 			y_counter += 65536;
 			
 			for (x=0; x < dst_width; x++) {
-				uintptr_t f = 0x80000000UL
-					/ dst_line1[x].weight;
-				*dst++ = (dst_line1[x].r * f) >> 15;
-				*dst++ = (dst_line1[x].g * f) >> 15;
-				*dst++ = (dst_line1[x].b * f) >> 15;
-				*dst++ = (dst_line1[x].a * f) >> 15;
+				uintptr_t f =  0x80000000UL / dst_line1[x].weight;
+				*dst++ = (val= (dst_line1[x].r * f) >> 15) > 255 ? 255: val;
+				*dst++ = (val= (dst_line1[x].g * f) >> 15) > 255 ? 255: val;
+				*dst++ = (val= (dst_line1[x].b * f) >> 15) > 255 ? 255: val;
+				*dst++ = (val= (dst_line1[x].a * f) >> 15) > 255 ? 255: val;
 			}
 			memset(dst_line1, 0, dst_width *
 			       sizeof(struct scale_outpix_byte));
@@ -667,13 +668,14 @@ static void shrink_picture_byte(
 		}
 	}
 	if (dst - dst_begin < dst_width * dst_height * 4) {
+		int val;
 		uintptr_t x;
 		for (x = 0; x < dst_width; x++) {
 			uintptr_t f = 0x80000000UL / dst_line1[x].weight;
-			*dst++ = (dst_line1[x].r * f) >> 15;
-			*dst++ = (dst_line1[x].g * f) >> 15;
-			*dst++ = (dst_line1[x].b * f) >> 15;
-			*dst++ = (dst_line1[x].a * f) >> 15;
+			*dst++ = (val= (dst_line1[x].r * f) >> 15) > 255 ? 255: val;
+			*dst++ = (val= (dst_line1[x].g * f) >> 15) > 255 ? 255: val;
+			*dst++ = (val= (dst_line1[x].b * f) >> 15) > 255 ? 255: val;
+			*dst++ = (val= (dst_line1[x].a * f) >> 15) > 255 ? 255: val;
 		}
 	}
 	MEM_freeN(dst_line1);
@@ -911,6 +913,8 @@ static void q_scale_float(float* in, float* out, int in_width,
    Should be comparable in speed to the ImBuf ..._fast functions at least 
    for byte-buffers.
 
+   NOTE: disabled, due to inacceptable inaccuracy and quality loss, see bug #18609 (ton)
+
 */
 static int q_scale_linear_interpolation(
 	struct ImBuf *ibuf, int newx, int newy)
@@ -1583,7 +1587,8 @@ struct ImBuf *IMB_scaleImBuf(struct ImBuf * ibuf, short newx, short newy)
 	scalefast_Z_ImBuf(ibuf, newx, newy);
 
 	/* try to scale common cases in a fast way */
-	if (q_scale_linear_interpolation(ibuf, newx, newy)) {
+	/* disabled, quality loss is inacceptable, see report #18609  (ton) */
+	if (0 && q_scale_linear_interpolation(ibuf, newx, newy)) {
 		return ibuf;
 	}