Sorry for the crosspost with the list:

Hello,

I bumped into an interesting problem with my third python SOP:

Basically I generate point positions for each integer frame, and from this, I (for now) linearly blend subframe data, which is fine.
When I eyeball my result in the viewport for non-integer frames, it all looks just fine.

Now, let's try motionblur as that's always the killer of this kind of things, directly rendering the SOP: nothing.

However, if I archive the geometry output of my SOP at $F and $F+shutter, and render that
(either as a DRA or simply with a fileSOP with no time*SOPs attached), I get the proper blur.
I tried this with all the three shutter evaluation modes, and (obviously after changing the archives) it all worked.

What am I missing? Or is this a limitation of PythonSOPs/Soho? A possible bug?

Thanks,
Andras