Page MenuHome

Continued incorrect single arm editing of aligned handles in graph editor
Closed, ResolvedPublic


Windows 7 Pro 64-bit
Intel Core i7-4770K CPU @3.50GHz
ASUS nVidia GeForce GTX 760 2GB GDDR5 DirectCU Mini

Broken: Blender 2.70 4849ca8

There is still incorrect behavior evident when attempting to edit a single curve control handle in the graph editor.

This incorrect behavior was reported in bug report T38594 and then in bug report T39947. In fact, it appears that the updates have introduced new problems in the editing behaviour.


PLEASE NOTE: I am listing these all together in one report because I believe that all of the incorrect behaviour mentioned below relates to the same bit of code and needs to be fixed all at the same time rather than bit by bit. If I am incorrect in this, let me know and I will happily break this down into separate bug reports.


The following problems are noted:

With the MOVE command 'G':

When a control arm is selected and moved, the arm opposite to the selected (and edited) control arm is now changing length, maintaining some sort of relationship with the edited control arm. This should not be happening.

In point of fact, this problem now makes it effectively impossible to edit the aligned control arms around an anchor point. If you move one arm, the other arm changes length. If you then select the other arm and attempt to move it back to the original length then the arm previously edited will now also change position.

NOTE: The use has to be able to move a single control arm of an aligned pair. When this happens the control arm on the opposite side will have to rotate to maintain alignment with the selected and edited arm. However, the unslected control arm should never – I repeat, NEVER – change length during this editing operation.

With the SCALE command 'S':

It is now impossible to scale a single arm of an aligned group around an anchor point in order to change its length (distance from the anchor point). Instead, both control arms – the selected one as well as the unselected one on the opposite side – scale in length together.

The user needs to be able to scale a single control arm in order to control the tension of the curve.

This was working correctly before the changes to this bit of code.

Additional editing problem with the SCALE command:

While executing the scaling operation the edited control arm can suddenly 'flip over', losing it's original arm angle, instead becoming completely vertical (see image).

This seems to have something to do with the X value of the arm length suddenly flipping over to a negative value. This is undesirable behaviour. The scaling values of the arm length should never be allowed to go negative during this operation.

[ Note: For the record, I have observed that the scaling values provided by the mouse cursor seem to be somewhat erratic and that input seems to vary depending upon just where the mouse cursor is in the editing window when the 'S' key is pressed to initiate the command, but that is something for another bug report. ]

With the ROTATE command 'R':

When rotating a single control arm around an anchor point, if that control arm reaches the end of its allowed rotation – becoming vertical in either a negative or positive orientation – and the mouse cursor is moved further then the length of the control arm suddenly enters into a scaling mode, allowing it to become shorter and even flip over the anchor point in the Y axis to the opposite orientation.

During the time that the control arm scales, the opposite control arm is also being allowed to scale in length. This is undesirable (see description of this problem above as related to the MOVE command)

This scale operation should not occur during a ROTATE command. If a control arm is rotated to the extreme vertical position (either positive or negative) then it should simply stop there and not be allowed to further nor to suddenly become shorter or flip over to negative.



Event Timeline

Ignatz (ignatz) added a project: BF Blender.
Ignatz (ignatz) set Type to Bug.
Ignatz (ignatz) added a subscriber: Ignatz (ignatz).
Ignatz (ignatz) created this task.
Ignatz (ignatz) raised the priority of this task from to Needs Triage by Developer.

@Joshua Leung (aligorith), while you're on looking into other issues with curves, mind looking into this one as well?

@Joshua Leung (aligorith) - Just wanted to add an extra observation about the ROTATE command.

[ Note: This observation applies to the code from Blender 2.69 but has bearing on the current situation ]

When only the anchor point is select (instead on a single control arm) then the rotatation occurs without scaling either of the arms attached to that anchor point.

On the other hand, if only a single anchor point is selected and then the rotate command is invoked, the arm on the opposite side of the anchor point is subject to wild overshoot.

This would seem to suggest that one should invoke that same anchor point ROTATE algorithm in those cases when only a single control arm is selected.

rBbc9e66f0830627e807f720dca4b9d5d8d39e732a reverted the change that broke scaling a single handle.

rotating is still a bit strange, but for now this is a known-limitation.

thanks for such a comprehensive report, but at this point the logic is changed so it matches 2.68 and previous.