FilletSrfToRail - Bug

FilletSrfToRail (nice new command !) fails here. (all red curves)
…also if the curve (red) is a former edge of filletSrf (green)

filletSrfRail-Fail.3dm (214.2 KB)

Version 8 WIP (8.0.23171.14306, 2023-06-20)
Expires August 4, 2023

1 Like

@pascal @Gijs
Pc-version is doing the job.
did you find time to check the mac ?

Hi Tom - I cannot get any result here.
RH-75632 FilletSrfToRail: failure case


1 Like

for me it looks like my nice new toy
is already broken on the mac side ?

1 Like

Version 8 WIP (8.0.23192.14306, 2023-07-11)

filletSrfToRail_fail2.3dm (3.3 MB)

another simple test-case - where filletSrfToRail fails
a cylinder with a tapered curve (ellipse).
full cylinder → fail completle
half cylinder → corrupt result
less then half → works

rebuilding the cylinder = workaround
@chuck maybe a startingpoint to solve this bug ?

1 Like

@pascal or @Gijs
looks like this is not solved.

Version 8 BETA (8.0.23286.01004, 2023-10-13)
Expires November 27, 2023

my guess is filletSrfToRail struggles with arcs

refering to this topic

I came along another example that does not work
after rebuilding the curve Degree 3, 10 CVs it works.
my guess: this bug is related to arcs / Degree 2 / Weight

see layer names for more info…

fillet issues_tp_rh8.3dm (4.5 MB)

would be nice to have this fixed for the V8 Release - thanks.

kind regards -tom

1 Like

just seen your comment on youtrack.
please have a look at above sample / fail - thanks

my guess: Circular stuff is not handled nicely…
maybe a starting point to solve the bug


1 Like

wait what :open_mouth:

@Tom_P Sorry, I’m just getting around to looking at this. FilletSrfToRail is quite a bit different from the other FilletSrf* things in that it doesn’t involve intersecting approximations to offsets to find a guess at the center curve and true the result up to the correct solution. It will have different types of problems. Here, it seems more like something is going wrong because only a small part of the curve will be involved in the fillet. I’ll dig deep and see if I can get it to work correctly. Thanks for the report.

I see that if you reverse the curve, the result looks pretty good. Must be something about where the curve starts relative to the second surface.

still sounds like alchemy.
we need rocket proven geometry …

thanks for having a look.

I think I have this one under control. Looks like the problem comes from the fact that the line from the left end of the curve to the lower left corner of the curved surface is tangent to both. This causes a numerical problem that was not anticipated. I have it working, but want to test some more.

1 Like

RH-77869 is fixed in Rhino 8 Service Release 2 Release Candidate

@Tom_P fyi I reopened the issue, as here it works in preview, but then the surface disappears when confirming.