However, the issue is difficult to produce. Here (https://www.screencast.com/t/cfMdBtDrOQ3) is a video showing the bug happening. You will notice that each time the item moves "drop" is logged. When a "cancel" event is emitted the nodes parent element must be the source container. You will notice that in the last "drop" a cancel event is emitted instead of a "drop" event even though the item has indead moved. The error can be identified in the event callback by a cancel event where the node's parent is not the source container.
I've been able to identify and reproduce a bug where, in some cases, a
dropis misidentified as acancel.A plunkr is here: https://embed.plnkr.co/9RBLky9DWbE2Inhp4sI1/.
However, the issue is difficult to produce. Here (https://www.screencast.com/t/cfMdBtDrOQ3) is a video showing the bug happening. You will notice that each time the item moves "drop" is logged. When a "cancel" event is emitted the nodes parent element must be the source container. You will notice that in the last "drop" a cancel event is emitted instead of a "drop" event even though the item has indead moved. The error can be identified in the event callback by a cancel event where the node's parent is not the source container.
A fix, and one we are using for now in @swimlane/ngx-dnd, is to check for this invalid state here: https://github.com/bevacqua/dragula/blob/master/dragula.js#L258 using something like
isInitialPlacement(target) && item.parentNode === target.