diff --git a/include/common.h b/include/common.h
index 5792f51..5c57326 100644
--- a/include/common.h
+++ b/include/common.h
@@ -19,7 +19,7 @@
 #define VERSION_X 0
 #define VERSION_Y 0
 #define VERSION_A 0
-#define VERSION_Z 7
+#define VERSION_Z 8
 // Remember to bump "Z" basically every time you change the engine!
 // TRIPLEFOX THIS INCLUDES YOU
 // Z can only be 0 for official releases!
diff --git a/pkg/base/common.lua b/pkg/base/common.lua
index ddd0cc4..6314035 100644
--- a/pkg/base/common.lua
+++ b/pkg/base/common.lua
@@ -82,7 +82,7 @@ MODE_TILT_SLOWDOWN = false -- TODO!
 MODE_TILT_DOWN_NOCLIMB = false -- TODO!
 
 MODE_DRUNKCAM_VELOCITY = false -- keep this off unless you want to throw up
-MODE_DRUNKCAM_LOCALTURN = true
+MODE_DRUNKCAM_LOCALTURN = false -- this is the one you're looking for.
 
 MODE_DELAY_SPADE_DIG = 1.0
 MODE_DELAY_SPADE_HIT = 0.25
diff --git a/pkg/base/obj_player.lua b/pkg/base/obj_player.lua
index de6ba7b..20f6971 100644
--- a/pkg/base/obj_player.lua
+++ b/pkg/base/obj_player.lua
@@ -699,6 +699,7 @@ function new_player(settings)
 			--print("polar",this.angx, this.angy)
 			
 		else
+			this.angx = this.angx + this.dangx
 			this.angy = this.angy + this.dangy
 		end
 		this.dangx = 0
diff --git a/pkg/base/version.lua b/pkg/base/version.lua
index 36f4e9d..5f60ac9 100644
--- a/pkg/base/version.lua
+++ b/pkg/base/version.lua
@@ -35,4 +35,5 @@ VERSION_BUGS = {
 {intro=4, fix=5, msg="Dedicated server build was broken"},
 {intro=5, fix=6, msg="CRASHES ON CHANNEL WRAPAROUND - PLEASE UPDATE TO 0.0-6!"},
 {intro=nil, fix=7, msg="Camera roll / camera_point_sky not implemented - drunken cam will not roll properly"},
+{intro=nil, fix=8, msg="Renderer uses double-rect approximation of cube instead of using trapezia"},
 }
diff --git a/pkg/iceball/pmfedit/main_client.lua b/pkg/iceball/pmfedit/main_client.lua
index d96b741..3da7dad 100644
--- a/pkg/iceball/pmfedit/main_client.lua
+++ b/pkg/iceball/pmfedit/main_client.lua
@@ -57,7 +57,9 @@ pmfedit_z = 0
 pmfedit_size = 16
 pmfedit_data = {}
 pmfedit_model = common.model_new(1)
-pmfedit_model, pmfedit_model_bone = common.model_bone_new(pmfedit_model)
+pmfedit_idx = 1
+pmfedit_model_bone = {}
+pmfedit_model, pmfedit_model_bone[pmfedit_idx] = common.model_bone_new(pmfedit_model)
 pmfedit_data[#pmfedit_data+1] = {x=0,y=0,z=0,r=0,g=0,b=0,radius=1}
 pmfedit_ry = 0
 pmfedit_rx = 0
@@ -110,7 +112,7 @@ function client.hook_tick(sec_current, sec_delta)
 		g=math.sin(sec_current)*127+128,
 		b=math.sin(sec_current+2*math.pi/3)*127+128,
 		radius=pmfedit_size}
-	common.model_bone_set(pmfedit_model, pmfedit_model_bone, "edit", pmfedit_data)
+	common.model_bone_set(pmfedit_model, pmfedit_model_bone[pmfedit_idx], "edit", pmfedit_data)
 
 	return 0.005
 end
@@ -213,9 +215,9 @@ function client.hook_key(key, state)
 			if xpmf then
 				common.model_free(pmfedit_model) -- YECCH! Forgot this line!
 				pmfedit_model = xpmf
-				pmfedit_model_bone = 0
+				pmfedit_model_bone[pmfedit_idx] = 0
 				local bname
-				bname, pmfedit_data = common.model_bone_get(pmfedit_model, pmfedit_model_bone)
+				bname, pmfedit_data = common.model_bone_get(pmfedit_model, pmfedit_model_bone[pmfedit_idx])
 				pmfedit_data[#pmfedit_data+1] = {}
 				print("loaded!")
 			else
@@ -225,8 +227,8 @@ function client.hook_key(key, state)
 			local xpt = pmfedit_data[#pmfedit_data]
 			pmfedit_data[#pmfedit_data] = nil
 			local bname, blah
-			bname, blah = common.model_bone_get(pmfedit_model, pmfedit_model_bone)
-			common.model_bone_set(pmfedit_model, pmfedit_model_bone, bname, pmfedit_data)
+			bname, blah = common.model_bone_get(pmfedit_model, pmfedit_model_bone[pmfedit_idx])
+			common.model_bone_set(pmfedit_model, pmfedit_model_bone[pmfedit_idx], bname, pmfedit_data)
 			if common.model_save_pmf(pmfedit_model, PMFEDIT_FNAME) then
 				print("saved!")
 			else
@@ -250,7 +252,7 @@ function client.hook_render()
 		,pmfedit_rx
 		,pmfedit_ry2))
 
-	client.model_render_bone_local(pmfedit_model, pmfedit_model_bone,
+	client.model_render_bone_local(pmfedit_model, pmfedit_model_bone[pmfedit_idx],
 		0,0,1,
 		pmfedit_ry,pmfedit_rx,pmfedit_ry2,
 		0.7)
@@ -261,9 +263,9 @@ if preload then
 	if xpmf then
 		common.model_free(pmfedit_model) -- YECCH! Forgot this line!
 		pmfedit_model = xpmf
-		pmfedit_model_bone = 0
+		pmfedit_model_bone[pmfedit_idx] = 0
 		local bname
-		bname, pmfedit_data = common.model_bone_get(pmfedit_model, pmfedit_model_bone)
+		bname, pmfedit_data = common.model_bone_get(pmfedit_model, pmfedit_model_bone[pmfedit_idx])
 		pmfedit_data[#pmfedit_data+1] = {}
 		print("loaded!")
 	else
diff --git a/src/render.c b/src/render.c
index 8c6e828..5e94a86 100644
--- a/src/render.c
+++ b/src/render.c
@@ -495,11 +495,147 @@ void render_vxl_rect_ftb_fast(uint32_t *ccolor, float *cdepth, int x1, int y1, i
 #endif
 }
 
+void render_vxl_cube_htrap(uint32_t *ccolor, float *cdepth,
+	int x1a, int x1b, int y1, int x2a, int x2b, int y2, uint32_t color, float depth)
+{
+	// dropout
+	if(x1b <= 0 && x2b <= 0)
+		return;
+	if(x1a >= cubemap_size && x2a >= cubemap_size)
+		return;
+	if(y2 <= 0)
+		return;
+	if(y1 >= cubemap_size)
+		return;
+	if(x1a >= x1b || x2a >= x2b)
+		return;
+	if(y1 >= y2)
+		return;
+	
+	// calc gradients
+	int m_x1a = (((x2a-x1a)<<16)+0x8000)/(y2-y1);
+	int m_x1b = (((x2b-x1b)<<16)+0x8000)/(y2-y1);
+	int sub_x1a = 0;
+	int sub_x1b = 0;
+	
+	// Y clamp
+	// TODO: clamp y1 properly
+	if(y2 >= cubemap_size)
+		y2 = cubemap_size;
+	
+	// render
+	//uint32_t *cptr = &ccolor[(y1<<cubemap_shift)+x1a];
+	//float *dptr = &cdepth[(y1<<cubemap_shift)+x1a];
+	uint32_t *cptr = ccolor;
+	float *dptr = cdepth;
+	int x,y;
+	for(y = y1; y < y2; y++)
+	{
+		if(y >= 0)
+		{
+			int rx1a = (x1a < 0 ? 0 : x1a);
+			int rx1b = (x1b >= cubemap_size ? cubemap_size : x1b);
+			
+			cptr = &ccolor[(y<<cubemap_shift)+rx1a];
+			dptr = &cdepth[(y<<cubemap_shift)+rx1a];
+			
+			for(x = rx1a; x < rx1b; x++)
+			{
+				if(*cptr == fog_color)
+				{
+					*cptr = color;
+					*dptr = depth;
+				}
+				cptr++;
+				dptr++;
+			}
+		}
+		
+		sub_x1a += m_x1a;
+		sub_x1b += m_x1b;
+		x1a += (sub_x1a>>16);
+		x1b += (sub_x1b>>16);
+		sub_x1a &= 0xFFFF;
+		sub_x1b &= 0xFFFF;
+	}
+}
+
+void render_vxl_cube_vtrap(uint32_t *ccolor, float *cdepth,
+	int x1, int y1a, int y1b, int x2, int y2a, int y2b, uint32_t color, float depth)
+{
+	// TODO: make this not so bloody horrible for the cache
+	
+	// dropout
+	if(y1b <= 0 && y2b <= 0)
+		return;
+	if(y1a >= cubemap_size && y2a >= cubemap_size)
+		return;
+	if(x2 <= 0)
+		return;
+	if(x1 >= cubemap_size)
+		return;
+	if(y1a >= y1b || y2a >= y2b)
+		return;
+	if(x1 >= x2)
+		return;
+	
+	// calc gradients
+	int m_y1a = ((y2a-y1a)<<16);
+	int m_y1b = ((y2b-y1b)<<16);
+	m_y1a /= (x2-x1);
+	m_y1b /= (x2-x1);
+	int sub_y1a = 0;
+	int sub_y1b = 0;
+	
+	// X clamp
+	// TODO: clamp x1 properly
+	// TODO: fix the leaks properly
+	x2++; y1b++; y2b++;
+	if(x2 >= cubemap_size)
+		x2 = cubemap_size;
+	
+	// render
+	//uint32_t *cptr = &ccolor[(y1<<cubemap_shift)+x1a];
+	//float *dptr = &cdepth[(y1<<cubemap_shift)+x1a];
+	uint32_t *cptr = ccolor;
+	float *dptr = cdepth;
+	int pitch = cubemap_size;
+	int x,y;
+	for(x = x1; x < x2; x++)
+	{
+		if(x >= 0)
+		{
+			int ry1a = (y1a < 0 ? 0 : y1a);
+			int ry1b = (y1b >= cubemap_size ? cubemap_size : y1b);
+			
+			cptr = &ccolor[(ry1a<<cubemap_shift)+x];
+			dptr = &cdepth[(ry1a<<cubemap_shift)+x];
+			
+			for(y = ry1a; y < ry1b; y++)
+			{
+				if(*cptr == fog_color)
+				{
+					*cptr = color;
+					*dptr = depth;
+				}
+				cptr += pitch;
+				dptr += pitch;
+			}
+		}
+		
+		sub_y1a += m_y1a;
+		sub_y1b += m_y1b;
+		y1a += (sub_y1a>>16);
+		y1b += (sub_y1b>>16);
+		sub_y1a &= 0xFFFF;
+		sub_y1b &= 0xFFFF;
+	}
+}
+
 void render_vxl_cube_sides(uint32_t *ccolor, float *cdepth, int x1, int y1, int x2, int y2, uint32_t color, float depth)
 {
 	int hsize = (cubemap_size>>1);
 	
-	
 	if(depth > CUBESUX_MARKER)
 	{
 		int x3 = ((x1-hsize)*depth)/(depth+1.0f)+hsize;
@@ -515,33 +651,32 @@ void render_vxl_cube_sides(uint32_t *ccolor, float *cdepth, int x1, int y1, int
 		return;
 	}
 	
-	int x3 = ((x1-hsize)*depth)/(depth+0.5f)+hsize;
-	int y3 = ((y1-hsize)*depth)/(depth+0.5f)+hsize;
-	int x4 = ((x2-hsize)*depth)/(depth+0.5f)+hsize;
-	int y4 = ((y2-hsize)*depth)/(depth+0.5f)+hsize;
+	int x3 = ((x1-hsize)*depth)/(depth+1.0f)+hsize;
+	int y3 = ((y1-hsize)*depth)/(depth+1.0f)+hsize;
+	int x4 = ((x2-hsize)*depth)/(depth+1.0f)+hsize;
+	int y4 = ((y2-hsize)*depth)/(depth+1.0f)+hsize+1;
 	
 	render_vxl_rect_ftb_fast(ccolor, cdepth, x1, y1, x2, y2, color, depth);
 	
 	depth += 0.5f;
 	
-	// TODO: replace these with trapezium drawing routines
-	if(x3 < x1)
-		render_vxl_rect_ftb_fast(ccolor, cdepth,
-			(int)x3, (int)y3, (int)x1, (int)y4,
-			color, depth);
-	else if(x2 < x4)
-		render_vxl_rect_ftb_fast(ccolor, cdepth,
-			(int)x2, (int)y3, (int)x4, (int)y4,
-			color, depth);
-	
 	if(y3 < y1)
-		render_vxl_rect_ftb_fast(ccolor, cdepth,
-			(int)x3, (int)y3, (int)x4, (int)y1,
-			color, depth);
+		render_vxl_cube_htrap(ccolor, cdepth,
+			x3, x4, y3, x1, x2, y1,
+			color, depth+1.0f);
 	else if(y2 < y4)
-		render_vxl_rect_ftb_fast(ccolor, cdepth,
-			(int)x3, (int)y2, (int)x4, (int)y4,
-			color, depth);
+		render_vxl_cube_htrap(ccolor, cdepth,
+			x1, x2, y2, x3, x4, y4,
+			color, depth+1.0f);
+	
+	if(x3 < x1)
+		render_vxl_cube_vtrap(ccolor, cdepth,
+			x3, y3, y4, x1, y1, y2,
+			color, depth+1.0f);
+	else if(x2 < x4)
+		render_vxl_cube_vtrap(ccolor, cdepth,
+			x2, y1, y2, x4, y3, y4,
+			color, depth+1.0f);
 }
 
 void render_vxl_cube(uint32_t *ccolor, float *cdepth, int x1, int y1, int x2, int y2, uint32_t color, float depth)