I’ve come across something with the Resample SOP that I’m not sure is expected behavior or possibly a bug.

I made a simple test (see attached screen capture) where I generate `curveu` three different ways after resampling a curve:

1. Calculating `curveu` directly on the same Resample SOP doing the resampling
2. Calculating `curveu` on a second Resample SOP downstream
3. Calculating `curveu` manually in a wrangle

The second and third approaches produce what I would expect — an evenly distributed parameterization along the resampled curve.

However, when `curveu` is generated on the same Resample SOP performing the resampling, the resulting values appear inaccurate or uneven.

This makes me wonder whether:

* this is an intentional behavior related to how the node evaluates internally,
* I’m misunderstanding how `curveu` is being generated in this context,
* or whether this could potentially be a bug.

Has anyone else encountered this or can explain the difference in behavior?
I have attached my test scene.

screen capture:
https://dl.dropboxusercontent.com/scl/fi/0mibffyivdvidzolp6f6w/houdini_hiNy1eWq9z.gif?rlkey=efrw40jc9g69y9t45igx4aqn7&dl=0 [dl.dropboxusercontent.com]

this is expected, even though not the most convenient
since Curve U Attribute output from Resample SOP will contain u value of the input curve not the output one

so while it's technically doing what the node help or tooltips describe, I tend to always use separate node dedicated to compute the u and in the domain I need, whether it is another Resample SOP, or vertexcurveparam() for evenly resampled curves or Edge Transport SOP or vertexcurveparam() + primuvconvert() combo for computing unitlength u domain on curves with uneven segment lengths, etc...

Yeah. The edge transport sop and resample for me. The problem is that the resample sop comes by default with resampling point toggle turned on. So people just think of as that sop that resamples points. But if you just turn it off and turn on the curve u attribute you just get the curve u attribute without resampling the curve.
The "distance along curve" node which is also an appropriate alias for the "edge transport"(Which is a more nerdy name) node can do the job as well