This was found using AMD-based CPUs with a higher thread count, such as the AMD Ryzen 9950X3D & AMD Ryzen Threadripper PRO 3975WX, 5955WX, and 5965WX CPUs. However, this has also been experienced with Intel CPUs such as i9-11900K @ 3.50GHz (16 CPUs)
Customer reported versionNuke 15.2v4
Customer reported platformWindows 11
Steps to reproduce1) In Nuke, create a Tracker node.
2) Select add track to create a new track.
3) Set the timeline frame to 100, then in the Viewer move the Track 1 anchor to create a keyframe animation.
4) Create a Roto node, and add 2 Bezier shapes.
5) Select the Root folder of the Roto1 node, then click on the Transform tab.
6) Open the Properties bin of the Tracker1 node, then hold CTRL + LMB click and drag the Transform squiggly to the Roto1 Transform xy squiggly to expression link them.
6b) Repeat the process for the any animated values of the Tracker1 to Roto1 knobs.
7) View the Roto node and press play (L) to see the animation play for the roto shapes. (continue playing)
8) In the Roto1 Properties bin, select the Motion blur tab.
9) Set the motion blur to global, then increase the motion blur knob to 4.
10) After the frames cache, stop the playback and observe the Node Graph for any label changes.
Expected behaviorThe node labels should be unaffected by motion blur properties of the linked Tracker and Roto nodes.
Actual behaviorAll nodes of the Nuke script will display an error on the label as: "cannot convert to a floating point value"
WorkaroundLimiting the CPU threads to 1 with the
-m 1 flag seemed to no longer experience the issue with the same steps to reproduce. More information on limiting threads in Nuke can be found here:
Q100117: Limiting threads in Nuke with the -m flag Reproduced by Support in: Nuke 16.0v6 - Windows 11 (Intel/AMD CPU)
Nuke 15.2v4 - Windows 11 (AMD CPU) -
regression Unable to reproduce bug in:
Nuke 16.0v6 - Windows 10, macOS Intel
Nuke 15.2v3 - Windows 11 (AMD CPU)
Earliest version tested15.2v3 - WIN11 AMD CPU- This issue doesn't appear in this version and has regressed