ID 470207 - Viewer scrubbing performance suffers when using hasError to update a node's Node Graph UI downstream of a heavy 3D animated scene

Problem summary
Viewer scrubbing performance suffers when using hasError to update a node's nodegraph UI downstream of a heavy 3D animated scene.

 

This happens when hasError is used on the label or disable knobs, which cause the node in the nodegraph to be updated visually, and happens with both the hasError TCL command or nuke.thisNode.hasError() Python command. This also happens with both ScanlineRender and RayRender nodes, and using a heavy non-3D nodes or a non-animated 3D scene upstream doesn't encounter the issue. If you reduce the poly count of the sphere, or the sample count of the render node, then this reduces the issue, increasing them make the issue worse. Playing the Viewer doesn't encounter the issue.

Turning lazy expression on or off makes no difference to the performance.

 

Customer reported version

nuke 11.3v6

 

Customer reported platform
Centos 7.6

 

Steps to reproduce

 

1) Open the attached Nuke script

 

​

2) While viewing the ColorWheel node, scrub the timeline using the mouse or arrow keys, Nuke is very slow to change frame
 

Expected behaviour
For the Viewer to change frame quickly when not Viewing a heavy script

 

Actual behaviour
Nuke takes a while to change frame, even when loading a animated ColorWheel

 

Workaround
If you group the node using hasError, then the issue stops happening.

 

Reproduced by Support in:

Nuke 12.2v5 - Windows 10, Centos 7.8

Nuke 12.2v1 - Windows 10

Nuke 12.0v1 - Windows 10

Nuke 11.0v1 - Windows 10

Nuke 10.0v1 - Windows 10

Nuke 9.0v1 - Windows 10, Centos 7.8

 

Earliest version tested

Nuke 9.0v1 - This issue appears to be in all versions of the product