<quiliro>
peeps[zen]: sorry... did not see your messages..
<quiliro>
the projection should not be a square from a piece of paper which hangs from opposite corners
<quiliro>
a square piece of paper
<quiliro>
my last code gets me a square projected...but it does not matter
<quiliro>
I just want to figure out which framework will stretch the paper in that way (without deforming)
<peeps[zen]>
well, its a catenary when viewed at 45degree angle, so to get one side of the frame, you would need to scale the input by sqrt(2), like catenary(x*sqrt(2)), but also I think you need to account for the length of that resulting curve to match the length of the paper
<peeps[zen]>
are you starting with a fixed square size?
<peeps[zen]>
or maybe divided by sqrt(2), not sure, just thinking out loud. it doesn't seem like there is a simple solution
<quiliro>
I could cut out another form. But it is easy to make a square, peeps[zen]
<quiliro>
the idea is to make a frame to hold a catenary whose projection is a square... or at least a figure with one right angle
<quiliro>
it could be a quarter of a circle
<quiliro>
that has one right angle
<quiliro>
and from the ends of the arc, a catenary could form
<quiliro>
but I guess the square is easier
<quiliro>
pulling it from opposite corners
<quiliro>
but not having all 4 corners on the same plane
<quiliro>
then, a catenary would form which would have its vertex on the lowest part of the sheet
califax has quit [Remote host closed the connection]
<peeps[zen]>
also, not sure how much it matters (depends on your goals i suppose), but keep in mind that a catenary is the shape formed by a handing rope/cable with a *uniform* weight distribution. the paper is wider at the bottom, but also stiffer, so maybe that cancels out
<peeps[zen]>
*hanging
califax has joined #openscad
<quiliro>
why is it stiffer?
<quiliro>
at the bottom
<quiliro>
it is wider at the center, but it will take the form of of the heaviest part: the longest catenary from corner to corner
<quiliro>
paper will not deform
<quiliro>
unless you pull too much or get it wet
<quiliro>
but then it will break
<peeps[zen]>
i just mean there is more material to bend at its widest, so forces are not uniformly distributed. i don't know, its more physics than I feel like thinking about right now
<quiliro>
what is wrong with this code:
<quiliro>
module pole {cylinder(2,.1,.1);}
<peeps[zen]>
best to explicitly name your parameters
<quiliro>
in line with what I have been taught here....
<quiliro>
peeps[zen]: how can I make a plane which is layed upon those poles?
<peeps[zen]>
i wouldn't call it a plane if it has creases or bends. planes are flat
<peeps[zen]>
anyways, i don't know
<quiliro>
how would you call it if it had catenaries, then?
<peeps[zen]>
just a surface i guess
<quiliro>
ok
<quiliro>
then I should describe it as: a plane which does not change dimensions but only direction when set on those 5 poles
<quiliro>
peeps[zen]: does that sound more correct?
<quiliro>
and stretched horizontally outside
<peeps[zen]>
a sheet that doesn't stretch, but does deform sagging
<quiliro>
ok
<quiliro>
dankon
<peeps[zen]>
seems like a problem similar to cloth simulation, which is typically done with FEM i think
LordOfBikes has quit [Ping timeout: 240 seconds]
<peeps[zen]>
i think blender might have some form of cloth simulation, but not sure (i only vaguely remember seeing maybe some screenshots or article about it some time ago). i don't really use blender though
arebil has quit [Quit: arebil]
<quiliro>
ok....probably blender does....but it is very difficult for me
<quiliro>
I would rather use omulas than drawing
<quiliro>
fomulas*
<quiliro>
formulas*
LordOfBikes has joined #openscad
<quiliro>
so the enunciation would be: a sheet that does not stretch, but does deform by sagging, is set on top of 4 poles which stand perpendicularly on a square and another (slightly taller) pole which is also standing perpendicularly, but on the center of the square
<quiliro>
so the enunciation would be: a sheet that does not stretch, but does deform by sagging, is set on top of 4 poles which stand perpendicularly on the corners of a square and on another (slightly taller) pole which is also standing perpendicularly, but on the center of the square
<quiliro>
sorry for the typo on the first version
teepee_ has joined #openscad
teepee has quit [Ping timeout: 264 seconds]
teepee_ is now known as teepee
mmu_man has quit [Ping timeout: 240 seconds]
quiliro has quit [Quit: ĝis!]
teepee has quit [Remote host closed the connection]
teepee has joined #openscad
teepee has quit [Remote host closed the connection]
<kintel>
Probably not many people awake at this point, but dropping this here: We should start putting together some requirements for moving Manifold out of experimental.
<kintel>
Some things to consider:
<kintel>
Co-existence of Manifold and CGAL (Nef polyhedrons). How to select which backend to use.
<kintel>
Missing Manifold features; How to structure that
<Scopeuk>
J23k55 YouCan try connecting a debugger to get some elements of stack etc. or there are some switches to make openscad chatty to a lot if I remember correctly --debug all having a quick look
<Scopeuk>
The debug argument should atleast give some idea what was happening at the time
<Scopeuk>
Not being able to reopen openscad is perculiarand suggests windows reused a memory allocation or at least something got cached
<J23k55>
ah command line parameter .. let me try this
<J23k55>
Scopeuk what is so weird is that openSCAD opens but works but super slow .. so opening takes a minute and every click or menu need at least 10sec
<J23k55>
hm ok i have no clue what this means Polygon2d-CGAL: Polygon2d::tessellate(): 2 outlines Polygon2d-CGAL: Polygon2d::tessellate(): 2 outlines
<J23k55>
oh now (without any change) i also get » terminate called after throwing an instance of 'std::out_of_range'
<J23k55>
what(): Vec out of range
<J23k55>
haha and on the 4th attempt (no changes just repeating the commandline) i finally got the 3mf rendered
<J23k55>
so i assume i hit a certain constellation where manifolds float precision causes a problem depending on the random tri calculation. This happened very seldom with 2023.08.13 - now with 2023.11.12 it is nearly every time when i use that module
RoyK has quit [Read error: Connection reset by peer]
RoyK has joined #openscad
mmu_man has joined #openscad
<InPhase>
J23k55: Speak to kintel. That is the only other time I heard of this absurd Windows effect where one needs to rename the executable to get it to work.
<InPhase>
J23k55: He spent many days looking into that.
<InPhase>
J23k55: Random thought I just had though. Does this behavior change if you turn off antivirus software?
<J23k55>
after some hours or reboot it also resets .. so windows is just keeping something - more important would be to find why manifold is causing this no more often
<J23k55>
no - but windows defender is always running something to copy your data
<J23k55>
maybe kintel knows what causes " terminate called after throwing an instance of 'std::out_of_range' what(): Vec out of range "
<J23k55>
At least the manifold errors shouldn't crash the software and just give some "mesh not closed " whatever error
<Scopeuk>
That sounds like something eliminated an element from a vector where something else has a cached size of, maybe we did a refinement operation out of sequence? Doesn't sound like something that should be tied 5o manifold 5hough
<Scopeuk>
Excuse 5's pretending to be T's stupid touch keyboards
<InPhase>
J23k55: Was that copy paste of the error?
<InPhase>
J23k55: Just making sure it was that and not "Vec: out of range"
<InPhase>
J23k55: i.e., colon vs no colon.
<InPhase>
And follow-up, if there is a colon, do you have an nvidia card?
erectus has quit [Remote host closed the connection]
erectus has joined #openscad
peepsalot has quit [Ping timeout: 260 seconds]
<J23k55>
InPhase yes nvidia card and i copied what was on the command line
kintel has joined #openscad
<kintel>
J23k55 Are you running OpenSCAD on the Window cmd-line?
<kintel>
My weird Windows executable rename issue was isolated to running on Windows on the cmd-line in a remote shell using a real GPU
<kintel>
Didn't figure out why, and haven't bothered reinstalling Windows yet
<kintel>
To get more debug output from OpenSCAD, try adding --debug=all on the cmd-line
<J23k55>
sometimes i also get this ( after scad is closed also on commandline) would say 3/10 tries https://imgur.com/a/nwbFv3I
<kintel>
right, looks like out of bounds read
<J23k55>
kintel both if i using the gui scad just closes so i took Scopeuk advice to use commandline with -debug all
<kintel>
If you could make a copy of the output whenever it happens that would be interesting
<kintel>
..and if you find a consistent way of making it fail, that would be helpful too
<kintel>
Anyway, I've seen two crash reports since we upgraded manifold a few days ago
<InPhase>
Manifold is at commit 96c57c059 from 2023-07-07
<J23k55>
the stl wasn't created (like the 3mf ) with the png i got pages of VertexArray: 168 : B glBufferSubData(GL_ARRAY_BUFFER, 1970160, 40, G?e@O@0þ]y?W?0>)
<InPhase>
kintel: Yeah, that's not there.
<InPhase>
kintel: There's a vec_dh.h in there, with different contents.
teepee has quit [Ping timeout: 264 seconds]
teepee_ is now known as teepee
<InPhase>
Now I could probably force this to update manifold, but before I do that, can we discuss what the right procedure should be for this to make sure we're using the same versions of the submodules?
<kintel>
commit b6eea0102
<kintel>
Tue Nov 7
<kintel>
I think git submodule update --init --recursive should do it
<kintel>
If not, do a git status and see if your manifold folder is dirty
<InPhase>
But are we actually supposed to run that before each build? Doesn't that say to update submodules past what upstream has?
<InPhase>
Because I don't want to be at latest, right? I want to be at what we've set the master branch as.
<kintel>
the update command just updates to whatever is in OpenSCAD
<kintel>
because git checkout doesn't pull from submodules
<InPhase>
Git git pull literally said "Fetching submodule submodules/manifold". Is it like lying to me?
<InPhase>
The manual update did update it.
<InPhase>
Before I did the update, git status said "nothing to commit, working tree clean" in manifold. After doing the git submodule update, now it says "Untracked files" and lists some random crap like pybind11, third_party/glm and thrust
<InPhase>
But I didn't touch anything. Somehow git made this unclean by updating?
<InPhase>
I'm not being crazy, right? This tool seems to be doing stupid things and not at all what's on the label. lol
Guest61 has joined #openscad
kintel has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Guest61 has quit [Client Quit]
<InPhase>
All I'm looking for is a ritual of commands to run that will let us all actually be working with the same source code. That's the one job of this tool. So should it be every time "git pull && git submodule update --init --recursive" from now on?
kintel has joined #openscad
<kintel>
right, it might be the --recursive that isn't done by default - I'm sure it's configurable though
<InPhase>
But specifically, do I need to run the submodule update step in addition to every pull just to keep the submodules pulled current?
<kintel>
that's what I do - haven't found a better way
<InPhase>
I keep a dedicated copy of the repository that I never edit, just so I have a local copy of the master branch in current state for code reading. So it's a little silly that I'm not having it actually current. :) But I can script this change.
<kintel>
..and the stray files are submodules pulled in by and older version of manifold itself
* InPhase
adds to the list of reasons someone needs to replace git.
<kintel>
I wish I had an idiot proof way of pulling myself, which doesn't require a clean clone :)
<InPhase>
Alright, now I have git_subpull script for pulling on all repositories with submodules...
<kintel>
J23k55 pages of debug output is expected :) ..but did it crash with STL or PNG export?
<kintel>
I'm wondering if it fails for the regular geometry ops, or if it's related to one of the file format exports, as there is a bunch of custom code for managing the geometry for those exports
<InPhase>
I had been sitting there searching for J23k55's error message, and oddly couldn't find it in the repository. Apparently that throw didn't exist in July.
<J23k55>
with STL yes same as 3mf .. sometimes it works but not always .. png was created 2/2 tries
<kintel>
J23k55 Are you building OpenSCAD yourself?
<J23k55>
no i using the dev snap from download section
<J23k55>
64bit
<kintel>
it's odd that this is a Windows-only issue, so I'm hoping for the Manifold folks to find it :)
<kintel>
..but none of their devs use Windows either : /
* J23k55
shouldn't use windows either but i was young and needed it for gaming
<J23k55>
so can confirm with png exports and render it also works only occasionally
<InPhase>
However, it's possible the problem existed before, but what was added was the bounds check.
<InPhase>
J23k55: So this is fully terminating OpenSCAD?
<J23k55>
"renamed VecDH into Vec" sounds fitting but i have no clue
<InPhase>
Because if so, the failure is actually our code, for not handling Manifold's exception throws.
<J23k55>
inPhase when using the gui it just vanishes .. with the command line it also terminates .. sometimes i catch a task running 100%
<InPhase>
If Manifold is throwing clean exceptions saying it had a problem, we should be converting those into user-displayed errors in the console. That's still a failure to get the render result, and still needs a fix, but we should not crash from this. All preview and render actions should be fully wrapped in a try/catch.
<InPhase>
We should have a big catch-all of std::exception& there and convert whatever we get to a displayed error.
<kintel>
InPhase good item for the final manifold qualification step
<kintel>
we have tons of code catching such CGAL errors..
kintel has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<InPhase>
J23k55: Can you post your error message as an issue, and include the note "preview and render should catch all generated exceptions at the top level" in the issue report?
<InPhase>
Or "catch and report"
<InPhase>
Because while manifold has something to fix there, so do we.
<InPhase>
Even if we catch and handle this more gracefully at a lower level, there should STILL be a top level to catch and report anything we miss.
<J23k55>
i can't find "open new issue" anymore
<InPhase>
lol
* InPhase
sighs at the world.
<InPhase>
J23k55: On the far right side.
<InPhase>
Did it used to be on the left?
<J23k55>
ah got it
<InPhase>
*thumbs-up*
J23k55 has quit [Quit: Ping timeout (120 seconds)]
J23k55 has joined #openscad
<InPhase>
And, sorry if I sounded short tempered in any of the above. I woke up with a bunch of sinus pain this morning, and I'm waiting for the medication to kick in fully. I recognized it might be making my tone a little more harsh, but I am attempting to mentally compensate and calibrate. :)
<teepee>
hmm, I suppose more annyoing to setup than actually hard. also potential difference of OpenGL dependent suff and mesh render due to mesa not liking our drawing code too much
<teepee>
maybe a reason to see if appveyor could help, the interactive support was quite useful