[Bf-blender-cvs] [3d243eb] master: Fix T49848: bevel of spiral gets bigger and bigger widths.

[Howard Trickey]

Commit: 3d243eb710dd7f2ddef3c421c1fd225390115185
Author: Howard Trickey
Date:   Wed Dec 28 13:55:19 2016 -0500
Branches: master
https://developer.blender.org/rB3d243eb710dd7f2ddef3c421c1fd225390115185

Fix T49848: bevel of spiral gets bigger and bigger widths.

This is the same issue as was fixed with T39486: the adjustment pass
that tries to equalize different widths at either end of an edge
sometimes causes the widths to get bigger and bigger.
The previous fix was to let "clamp_overlap" do double duty as a way
to limit this behavior. But clearly this is undiscoverable, as the
current bug report shows.  So I put in an "auto-limiting" mode that
detects when adjustments are going crazy and then acts as if
clamp_overlap were set.
The reason we can't always act as if clamp_overlap is set is that
certain models (e.g., Bent_test in regression tests) look bad if
that is enabled.

===================================================================

M	source/blender/bmesh/tools/bmesh_bevel.c

===================================================================

diff --git a/source/blender/bmesh/tools/bmesh_bevel.c b/source/blender/bmesh/tools/bmesh_bevel.c
index 8340be8..05169a2 100644
--- a/source/blender/bmesh/tools/bmesh_bevel.c
+++ b/source/blender/bmesh/tools/bmesh_bevel.c
@@ -57,6 +57,7 @@
 #define BEVEL_EPSILON_ANG DEG2RADF(2.0f)
 #define BEVEL_SMALL_ANG DEG2RADF(10.0f)
 #define BEVEL_MAX_ADJUST_PCT 10.0f
+#define BEVEL_MAX_AUTO_ADJUST_PCT 300.0f
 
 /* happens far too often, uncomment for development */
 // #define BEVEL_ASSERT_PROJECT
@@ -323,15 +324,33 @@ static bool edge_half_offset_changed(EdgeHalf *e)
 	       e->offset_r != e->offset_r_spec;
 }
 
-static bool any_edge_half_offset_changed(BevVert *bv)
+static float adjusted_rel_change(float val, float spec)
+{
+	float relchg;
+
+	relchg = 0.0f;
+	if (val != spec) {
+		if (spec == 0)
+			relchg = 1000.0f;  /* arbitrary large value */
+		else
+			relchg = fabsf((val - spec) / spec);
+	}
+	return relchg;
+}
+
+static float max_edge_half_offset_rel_change(BevVert *bv)
 {
 	int i;
+	float max_rel_change;
+	EdgeHalf *e;
 
+	max_rel_change = 0.0f;
 	for (i = 0; i < bv->edgecount; i++) {
-		if (edge_half_offset_changed(&bv->edges[i]))
-			return true;
+		e = &bv->edges[i];
+		max_rel_change = max_ff(max_rel_change, adjusted_rel_change(e->offset_l, e->offset_l_spec));
+		max_rel_change = max_ff(max_rel_change, adjusted_rel_change(e->offset_r, e->offset_r_spec));
 	}
-	return false;
+	return max_rel_change;
 }
 
 /* Return the next EdgeHalf after from_e that is beveled.
@@ -1951,6 +1970,7 @@ static void adjust_offsets(BevelParams *bp)
 	GHashIterator giter;
 	EdgeHalf *e, *efirst, *eother;
 	GSQueue *q;
+	float max_rel_adj;
 
 	BLI_assert(!bp->vertex_only);
 	GHASH_ITER(giter, bp->vert_hash) {
@@ -1966,7 +1986,7 @@ static void adjust_offsets(BevelParams *bp)
 		searchi = -1;
 		GHASH_ITER(giter, bp->vert_hash) {
 			bv = BLI_ghashIterator_getValue(&giter);
-			if (!bv->visited && any_edge_half_offset_changed(bv)) {
+			if (!bv->visited && max_edge_half_offset_rel_change(bv) > 0.0f) {
 				i = BM_elem_index_get(bv->v);
 				if (!searchbv || i < searchi) {
 					searchbv = bv;
@@ -1996,6 +2016,24 @@ static void adjust_offsets(BevelParams *bp)
 		}
 	}
 	BLI_gsqueue_free(q);
+
+	/* Should we auto-limit the error accumulation? Typically, spirals can lead to 100x relative adjustments,
+	 * and somewhat hacky mechanism of using bp->limit_offset to indicate "clamp the adjustments" is not
+	 * obvious to users, who almost certainaly want clamping in this situation.
+	 * The reason not to clamp always is that some models work better without it (e.g., Bent_test in regression
+	 * suite, where relative adjust maximum is about .6). */
+	if (!bp->limit_offset) {
+		max_rel_adj = 0.0f;
+		GHASH_ITER(giter, bp->vert_hash) {
+			bv = BLI_ghashIterator_getValue(&giter);
+			max_rel_adj = max_ff(max_rel_adj, max_edge_half_offset_rel_change(bv));
+		}
+		if (max_rel_adj > BEVEL_MAX_AUTO_ADJUST_PCT / 100.0f) {
+			bp->limit_offset = true;
+			adjust_offsets(bp);
+			bp->limit_offset = false;
+		}
+	}
 }
 
 /* Do the edges at bv form a "pipe"?