Array Modifier "Merge" bug? #41983

New Issue

Admiral Potato · 2014-09-28T03:21:27+02:00

Admiral Potato commented

2014-09-28 03:21:27 +02:00

System Information
Windows 7 Ultimate, 64 bit, GTX 680
Windows 8 pro, GTX 780ti

Blender Version
Broken: 2.72 RC1
Worked: 2.71

Short description of error
It seems that the Array modifier's "Merge" feature is no longer correctly. It seems to skip over connecting every other generated copy?

Exact steps for others to reproduce the error
In the attached .blend file, take a look at the top of this Octopus in both 2.71 and 2.72. The effect of the breakage seems to be exaggerated when using "alt-a" to play back the animation in an OpenGL shading mode.

volume_octopus3_bugreport.blend

(simple example of bug array_mod.blend)

**System Information** Windows 7 Ultimate, 64 bit, GTX 680 Windows 8 pro, GTX 780ti **Blender Version** Broken: 2.72 RC1 Worked: 2.71 **Short description of error** It seems that the Array modifier's "Merge" feature is no longer correctly. It seems to skip over connecting every other generated copy? **Exact steps for others to reproduce the error** In the attached .blend file, take a look at the top of this Octopus in both 2.71 and 2.72. The effect of the breakage seems to be exaggerated when using "alt-a" to play back the animation in an OpenGL shading mode. ![8-adaptive_256-diff_100.gif](https://archive.blender.org/developer/F113275/8-adaptive_256-diff_100.gif) [volume_octopus3_bugreport.blend](https://archive.blender.org/developer/F113274/volume_octopus3_bugreport.blend) (simple example of bug [array_mod.blend](https://archive.blender.org/developer/F114106/array_mod.blend))

Admiral Potato commented

2014-09-28 03:21:27 +02:00

Changed status to: 'Open'

Admiral Potato commented

2014-09-28 03:21:27 +02:00

Added subscriber: @AdmiralPotato

blender-admin commented

2014-09-28 03:21:27 +02:00

#42031 was marked as duplicate of this issue

Bastien Montagne commented

2014-09-28 08:46:58 +02:00

Added subscriber: @mont29

Bastien Montagne self-assigned this 2014-09-28 08:47:06 +02:00

Bastien Montagne commented

2014-09-28 17:38:10 +02:00

Added subscriber: @ideasman42

Bastien Montagne commented

2014-09-28 17:38:10 +02:00

Gah, this behavior was intended, from comments in code. Yet, I really see no good reason for it - if it’s the idea that this could end in merging vertices that are farther than threshold distance (neighbor-from-neighbor additive 'error'), it’s way less painful than bug reported here imho.

Campbell, really need your advice here, this simple patch fixes the bug, do you see any issue with it?

P149: (An Untitled Masterwork)

diff --git a/source/blender/modifiers/intern/MOD_array.c b/source/blender/modifiers/intern/MOD_array.c
index 6679648..d338ac0 100644
--- a/source/blender/modifiers/intern/MOD_array.c
+++ b/source/blender/modifiers/intern/MOD_array.c
@@ -606,11 +606,6 @@ static DerivedMesh *arrayModifier_doArray(
 					int target = full_doubles_map[prev_chunk_index];
 					if (target != -1) {
 						target += chunk_nverts; /* translate mapping */
-						/* The rule here is to not follow mapping to chunk N-2, which could be too far
-						 * so if target vertex was itself mapped, then this vertex is not mapped */
-						if (full_doubles_map[target] != -1) {
-							target = -1;
-						}
 					}
 					full_doubles_map[this_chunk_index] = target;
 				}
@@ -624,7 +619,7 @@ static DerivedMesh *arrayModifier_doArray(
 				        c * chunk_nverts,
 				        chunk_nverts,
 				        amd->merge_dist,
-				        false);
+				        true);
 			}
 		}
 	}
@@ -644,7 +639,7 @@ static DerivedMesh *arrayModifier_doArray(
 		        first_chunk_start,
 		        first_chunk_nverts,
 		        amd->merge_dist,
-		        false);
+		        true);
 	}
 
 	/* start capping */
@@ -669,7 +664,7 @@ static DerivedMesh *arrayModifier_doArray(
 			        start_cap_start,
 			        start_cap_nverts,
 			        amd->merge_dist,
-			        false);
+			        true);
 		}
 	}
 
@@ -694,7 +689,7 @@ static DerivedMesh *arrayModifier_doArray(
 			        end_cap_start,
 			        end_cap_nverts,
 			        amd->merge_dist,
-			        false);
+			        true);
 		}
 	}
 	/* done capping */
@@ -711,7 +706,12 @@ static DerivedMesh *arrayModifier_doArray(
 	if (use_merge) {
 		for (i = 0; i < result_nverts; i++) {
 			if (full_doubles_map- [x] != -1) {
-				tot_doubles++;
+				if (i == full_doubles_map[i]) {
+					full_doubles_map- [x] = -1;
+				}
+				else {
+					tot_doubles++;
+				}
 			}
 		}
 		if (tot_doubles > 0) {

Gah, this behavior was intended, from comments in code. Yet, I really see no good reason for it - if it’s the idea that this could end in merging vertices that are farther than threshold distance (neighbor-from-neighbor additive 'error'), it’s way less painful than bug reported here imho. Campbell, really need your advice here, this simple patch fixes the bug, do you see any issue with it? [P149: (An Untitled Masterwork)](https://archive.blender.org/developer/P149.txt) ``` diff --git a/source/blender/modifiers/intern/MOD_array.c b/source/blender/modifiers/intern/MOD_array.c index 6679648..d338ac0 100644 --- a/source/blender/modifiers/intern/MOD_array.c +++ b/source/blender/modifiers/intern/MOD_array.c @@ -606,11 +606,6 @@ static DerivedMesh *arrayModifier_doArray( int target = full_doubles_map[prev_chunk_index]; if (target != -1) { target += chunk_nverts; /* translate mapping */ - /* The rule here is to not follow mapping to chunk N-2, which could be too far - * so if target vertex was itself mapped, then this vertex is not mapped */ - if (full_doubles_map[target] != -1) { - target = -1; - } } full_doubles_map[this_chunk_index] = target; } @@ -624,7 +619,7 @@ static DerivedMesh *arrayModifier_doArray( c * chunk_nverts, chunk_nverts, amd->merge_dist, - false); + true); } } } @@ -644,7 +639,7 @@ static DerivedMesh *arrayModifier_doArray( first_chunk_start, first_chunk_nverts, amd->merge_dist, - false); + true); } /* start capping */ @@ -669,7 +664,7 @@ static DerivedMesh *arrayModifier_doArray( start_cap_start, start_cap_nverts, amd->merge_dist, - false); + true); } } @@ -694,7 +689,7 @@ static DerivedMesh *arrayModifier_doArray( end_cap_start, end_cap_nverts, amd->merge_dist, - false); + true); } } /* done capping */ @@ -711,7 +706,12 @@ static DerivedMesh *arrayModifier_doArray( if (use_merge) { for (i = 0; i < result_nverts; i++) { if (full_doubles_map- [x] != -1) { - tot_doubles++; + if (i == full_doubles_map[i]) { + full_doubles_map- [x] = -1; + } + else { + tot_doubles++; + } } } if (tot_doubles > 0) { ```

Sebastian Włoszek commented

2014-09-29 12:20:09 +02:00

Added subscriber: @Idlero

Bastien Montagne commented

2014-09-29 12:21:06 +02:00

Think this needs the regression tag, in fact :|

Bastien Montagne commented

2014-10-01 07:23:04 +02:00

Added subscriber: @igvalor2

Igor Valorin commented

2014-10-01 13:20:53 +02:00

So... Will the issue be resolved?

Campbell Barton commented

2014-10-02 11:04:29 +02:00

This comment was removed by @ideasman42

*This comment was removed by @ideasman42*

blender-admin commented

2014-10-03 14:36:34 +02:00

This issue was referenced by 4c43fcf791

This issue was referenced by 4c43fcf791c3c26fe5f729ffe5e96bae2957e88a

Campbell Barton commented

2014-10-03 14:41:47 +02:00

Changed status from 'Open' to: 'Resolved'

Campbell Barton closed this issue

2014-10-03 14:41:47 +02:00

Campbell Barton commented

2014-10-03 14:41:47 +02:00

Closed by commit 4c43fcf791.

Closed by commit 4c43fcf791.

blender-admin commented

2014-10-03 15:09:32 +02:00

This issue was referenced by 561b52b356

This issue was referenced by 561b52b3562ab4e580d67cb228bd0515427af6ef

Admiral Potato commented

2014-10-05 20:47:42 +02:00

Changed status from 'Resolved' to: 'Open'

Admiral Potato reopened this issue

2014-10-05 20:47:42 +02:00

Admiral Potato commented

2014-10-05 20:47:42 +02:00

Sadly, when I opened the blend file that I attached to this bug report with the newly released 2.72, I still had a similar but slightly different issue when animating. It now looks like the modifier is trying to find the very first vertex in the mesh, and merge all of the radially merged verts with that one, but only every other frame of animation. Would you please take a look at the previously attached blend file once more, and press alt-a to start animating it? That's where the problem persists.

See attached animated GIF for an example.

Sadly, when I opened the blend file that I attached to this bug report with the newly released 2.72, I still had a similar but slightly different issue when animating. It now looks like the modifier is trying to find the very first vertex in the mesh, and merge all of the radially merged verts with that one, but only every other frame of animation. Would you please take a look at the previously attached blend file once more, and press alt-a to start animating it? That's where the problem persists. See attached animated GIF for an example. ![b0rken_at_272_realease.gif](https://archive.blender.org/developer/F115155/b0rken_at_272_realease.gif)

blender-admin commented

2014-10-05 23:10:44 +02:00

This issue was referenced by 1dd428ff6d

This issue was referenced by 1dd428ff6d45194fcabe1bdff62ede08f531a4ea

Bastien Montagne commented

2014-10-05 23:12:09 +02:00

Changed status from 'Open' to: 'Resolved'

Bastien Montagne closed this issue

2014-10-05 23:12:09 +02:00

Bastien Montagne commented

2014-10-05 23:12:09 +02:00

Closed by commit 1dd428ff6d.