failing preview and malformed full render
I've just cloned/compiled and built the latest 3rd September 2025
OpenSCAD code, in the hope that issues with previous builds [after
2021.01, which works OK] have been resolved, They haven't :-(. I also
want to try out the namespace code..
OpenSCAD install built directly from a git clone and following the
[perfect] instructions for a Debian 12 desktop.
The original code is from http://www.thingiverse.com/thing:3575,
The version I use, [lightly modified to create the object I want] is
attached.
Preview always renders with CSG, painfully slow of course, and the main
gear object does not preview with teeth, and on moving the Viewport
around, preview shows a truncated cone when the +Z axis is up [I
normally view this with +Z down].
The ring that appears to render in fresh air actually serves as a prop
for the corners of the teeth when exported as an STL for printing, the
gear is printed coned side up as that is the visible side, there's a
second ring underneath, which breaks up the support into sections so
that when printed the rings and support can be removed more easily.
Performing a full render results in clearly visible gaps and slot-like
grooves on the teeth, which are reproduced in an export, making the STL
unusable.
Could the development team comment please?
Much appreciated.
Roger.
On 9/3/2025 5:52 AM, Roger Whiteley via Discuss wrote:
Could the development team comment please?
I don't know whether I qualify as a member of the development team - I
haven't gotten my membership card yet - but my comment is "indeed,
that's interesting; perhaps you could trim it down to a more minimal
demonstration case".
Everybody: please don't ask the development team to pull apart a
multi-hundred line program to figure out where the problem is. Do what
you can first. Chop away everything that doesn't demonstrate the
problem; normally you can get it down to only a dozen or two lines.
Doing that doesn't require any knowledge of the internals, but does
require knowledge of your program - which you have, and the developer
looking at it doesn't have. Often, in the process you'll find a mistake
in the program.
Here it's pretty clear that there is a problem in OpenSCAD that leads
to the preview being different from the render, but you could still chop
it down to the minimum demonstration. The notches are probably an
OpenSCAD issue too, since they are there with Manifold but not with CGAL.
Could it be that the object making the holes is exactly coincident with the
top of the hole. GCAL will handle that if it is exactly coincident, but
perhaps manifold depends on floating point accuracy.
On Wed, 3 Sept 2025 at 18:25, Jordan Brown via Discuss <
discuss@lists.openscad.org> wrote:
On 9/3/2025 5:52 AM, Roger Whiteley via Discuss wrote:
Could the development team comment please?
I don't know whether I qualify as a member of the development team - I
haven't gotten my membership card yet - but my comment is "indeed, that's
interesting; perhaps you could trim it down to a more minimal demonstration
case".
Everybody: please don't ask the development team to pull apart a
multi-hundred line program to figure out where the problem is. Do what you
can first. Chop away everything that doesn't demonstrate the problem;
normally you can get it down to only a dozen or two lines. Doing that
doesn't require any knowledge of the internals, but does require knowledge
of your program - which you have, and the developer looking at it doesn't
have. Often, in the process you'll find a mistake in the program.
Here it's pretty clear that there is a problem in OpenSCAD that leads to
the preview being different from the render, but you could still chop it
down to the minimum demonstration. The notches are probably an OpenSCAD
issue too, since they are there with Manifold but not with CGAL.
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
We haven't yet have issues with that. The problem openscad usually have
(not necessarily this issue) is when converting manifold mesh to CGAL Nef
polyhedron for minkowski operation.
On Thu, Sep 4, 2025, 01:58 nop head via Discuss discuss@lists.openscad.org
wrote:
Could it be that the object making the holes is exactly coincident with
the top of the hole. GCAL will handle that if it is exactly coincident, but
perhaps manifold depends on floating point accuracy.
On Wed, 3 Sept 2025 at 18:25, Jordan Brown via Discuss <
discuss@lists.openscad.org> wrote:
On 9/3/2025 5:52 AM, Roger Whiteley via Discuss wrote:
Could the development team comment please?
I don't know whether I qualify as a member of the development team - I
haven't gotten my membership card yet - but my comment is "indeed, that's
interesting; perhaps you could trim it down to a more minimal demonstration
case".
Everybody: please don't ask the development team to pull apart a
multi-hundred line program to figure out where the problem is. Do what you
can first. Chop away everything that doesn't demonstrate the problem;
normally you can get it down to only a dozen or two lines. Doing that
doesn't require any knowledge of the internals, but does require knowledge
of your program - which you have, and the developer looking at it doesn't
have. Often, in the process you'll find a mistake in the program.
Here it's pretty clear that there is a problem in OpenSCAD that leads
to the preview being different from the render, but you could still chop it
down to the minimum demonstration. The notches are probably an OpenSCAD
issue too, since they are there with Manifold but not with CGAL.
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
Jordan
I should have been more clear that I wasn't expecting to get someone
else to debug the code, only that this code causes the preview and
rendering systems to not generate a clean object.
Thanks for the suggestion, this is one of the problems with a code block
that I am essentially reusing without necessarily knowing exactly how it
functions :-(, not always a good starting point.
I could look through other code that I own and understand and try to
find a simpler use case, I do recall issues with something else,
particularly in Preview, now I've built the latest release, its
installed right over the 2021.05 official version, I still have the
2024.12.30 version installed via Flatpak, but I think reinstalling from
the Debian repo will tread over the latest build, such is life..
I've taken a look at the code as it stands, and removed two assign
statements, @ lines 529 and 533, that I missed, and removed extra braces
that performed no function.
So I've (re)downloaded the original, and sure enough there are other
assign statements which I changed to regular variable = some value.
The original still fails to preview correctly, leaving all the teeth
off, making it a good variable drive system, but not much good for
transmitting power..
There is a much newer library of bevel gears, here
https://www.thingiverse.com/thing:6881272, which I will have a play
with, rendering looks good though.
It would be good to simply understand if there's a repeatable problem
with other code.
Roger.
On 03/09/2025 18:24, Jordan Brown wrote:
On 9/3/2025 5:52 AM, Roger Whiteley via Discuss wrote:
Could the development team comment please?
I don't know whether I qualify as a member of the development team - I
haven't gotten my membership card yet - but my comment is "indeed,
that's interesting; perhaps you could trim it down to a more minimal
demonstration case".
Everybody: please don't ask the development team to pull apart a
multi-hundred line program to figure out where the problem is. Do
what you can first. Chop away everything that doesn't demonstrate the
problem; normally you can get it down to only a dozen or two lines.
Doing that doesn't require any knowledge of the internals, but does
require knowledge of your program - which you have, and the developer
looking at it doesn't have. Often, in the process you'll find a
mistake in the program.
Here it's pretty clear that there is a problem in OpenSCAD that
leads to the preview being different from the render, but you could
still chop it down to the minimum demonstration. The notches are
probably an OpenSCAD issue too, since they are there with Manifold but
not with CGAL.