[Bf-blender-cvs] SVN commit: /data/svn/bf-blender [13592] trunk/blender/source/blender/src/ transform.c:

[Brecht Van Lommel]

Revision: 13592
          http://projects.blender.org/plugins/scmsvn/viewcvs.php?view=rev&root=bf-blender&revision=13592
Author:   blendix
Date:     2008-02-06 18:29:16 +0100 (Wed, 06 Feb 2008)

Log Message:
-----------

Bugfix for closely zoomed in rotation transform, with the rotation
center far out of the view, gave integer overflow and wrong results,
now uses doubles to avoid the overflow.

Modified Paths:
--------------
    trunk/blender/source/blender/src/transform.c

Modified: trunk/blender/source/blender/src/transform.c
===================================================================
--- trunk/blender/source/blender/src/transform.c	2008-02-06 16:45:51 UTC (rev 13591)
+++ trunk/blender/source/blender/src/transform.c	2008-02-06 17:29:16 UTC (rev 13592)
@@ -3088,16 +3088,16 @@
 
 	float final;
 
-	int dx2 = t->center2d[0] - mval[0];
-	int dy2 = t->center2d[1] - mval[1];
+	double dx2 = t->center2d[0] - mval[0];
+	double dy2 = t->center2d[1] - mval[1];
 	double B = sqrt(dx2*dx2+dy2*dy2);
 
-	int dx1 = t->center2d[0] - t->imval[0];
-	int dy1 = t->center2d[1] - t->imval[1];
+	double dx1 = t->center2d[0] - t->imval[0];
+	double dy1 = t->center2d[1] - t->imval[1];
 	double A = sqrt(dx1*dx1+dy1*dy1);
 
-	int dx3 = mval[0] - t->imval[0];
-	int dy3 = mval[1] - t->imval[1];
+	double dx3 = mval[0] - t->imval[0];
+	double dy3 = mval[1] - t->imval[1];
 		/* use doubles here, to make sure a "1.0" (no rotation) doesnt become 9.999999e-01, which gives 0.02 for acos */
 	double deler= ((double)((dx1*dx1+dy1*dy1)+(dx2*dx2+dy2*dy2)-(dx3*dx3+dy3*dy3) ))
 		/ (2.0 * (A*B?A*B:1.0));
@@ -3586,23 +3586,23 @@
 
 	float final;
 
-	int dx2 = t->center2d[0] - mval[0];
-	int dy2 = t->center2d[1] - mval[1];
-	float B = (float)sqrt(dx2*dx2+dy2*dy2);
+	double dx2 = t->center2d[0] - mval[0];
+	double dy2 = t->center2d[1] - mval[1];
+	double B = (float)sqrt(dx2*dx2+dy2*dy2);
 
-	int dx1 = t->center2d[0] - t->imval[0];
-	int dy1 = t->center2d[1] - t->imval[1];
-	float A = (float)sqrt(dx1*dx1+dy1*dy1);
+	double dx1 = t->center2d[0] - t->imval[0];
+	double dy1 = t->center2d[1] - t->imval[1];
+	double A = (float)sqrt(dx1*dx1+dy1*dy1);
 
-	int dx3 = mval[0] - t->imval[0];
-	int dy3 = mval[1] - t->imval[1];
+	double dx3 = mval[0] - t->imval[0];
+	double dy3 = mval[1] - t->imval[1];
 
-	float deler= ((dx1*dx1+dy1*dy1)+(dx2*dx2+dy2*dy2)-(dx3*dx3+dy3*dy3))
+	double deler= ((dx1*dx1+dy1*dy1)+(dx2*dx2+dy2*dy2)-(dx3*dx3+dy3*dy3))
 		/ (2 * A * B);
 
 	float dphi;
 
-	dphi = saacos(deler);
+	dphi = saacos((float)deler);
 	if( (dx1*dy2-dx2*dy1)>0.0 ) dphi= -dphi;
 
 	if(G.qual & LR_SHIFTKEY) t->fac += dphi/30.0f;
@@ -4183,16 +4183,16 @@
 
 	float final;
 
-	int dx2 = t->center2d[0] - mval[0];
-	int dy2 = t->center2d[1] - mval[1];
+	double dx2 = t->center2d[0] - mval[0];
+	double dy2 = t->center2d[1] - mval[1];
 	double B = sqrt(dx2*dx2+dy2*dy2);
 
-	int dx1 = t->center2d[0] - t->imval[0];
-	int dy1 = t->center2d[1] - t->imval[1];
+	double dx1 = t->center2d[0] - t->imval[0];
+	double dy1 = t->center2d[1] - t->imval[1];
 	double A = sqrt(dx1*dx1+dy1*dy1);
 
-	int dx3 = mval[0] - t->imval[0];
-	int dy3 = mval[1] - t->imval[1];
+	double dx3 = mval[0] - t->imval[0];
+	double dy3 = mval[1] - t->imval[1];
 		/* use doubles here, to make sure a "1.0" (no rotation) doesnt become 9.999999e-01, which gives 0.02 for acos */
 	double deler= ((double)((dx1*dx1+dy1*dy1)+(dx2*dx2+dy2*dy2)-(dx3*dx3+dy3*dy3) ))
 		/ (2.0 * (A*B?A*B:1.0));