<teepee>
not sure if it can do just any mount, there is something for removable devices but that looks like targeting actual usb-sticks and such
<teepee>
anyway, past time to get some sleep, maybe I'll find a minute to look through the docs tomorrow
<stealth_>
i just uninstalled the snap package and start using AppImage, works fine, less headache
<teepee>
that can have other issues, but it's not a container, so no file system access dances ;-)
<stealth_>
no worries.
<stealth_>
questio: say i have a cylinder and i move it `x=1`, the position is set not on 1 but offset by something? if the `cylinter(center=true)` then it seems to center. is there position you can use to set left/right/top-right, ...?
<stealth_>
nm i suppose you just use translate() to offset as you need it.
hisacro has quit [Ping timeout: 264 seconds]
hisacro has joined #openscad
snaked has quit [Ping timeout: 260 seconds]
J24k10 has joined #openscad
J24k15 has quit [Ping timeout: 256 seconds]
mmu_man has quit [Ping timeout: 252 seconds]
snaked has joined #openscad
ferdna has joined #openscad
jaredo has quit [*.net *.split]
jaredo has joined #openscad
marcus has quit [Ping timeout: 248 seconds]
marcus has joined #openscad
misterfish has joined #openscad
GNUmoon has quit [Remote host closed the connection]
GNUmoon has joined #openscad
<J24k10>
so the 2D systems causes SCAD to close unexpected if offsets are applied that have exactly size/2 of the object (so the object or gap would vanish)
<J24k10>
(happens also without manifold)
JakeSays has joined #openscad
JakeSays_ has quit [Ping timeout: 252 seconds]
<stealth_>
J24k10, if its a simple script i can run it on my system see if it happens as well.
<J24k10>
it is voronoi shape that is in my librarie
<J24k10>
Rund(.5)Voronoi(30,dicke=1);
<J24k10>
https://imgur.com/a/hFimkvm this is the shape and substract and apply an offset of .5 (the gaps are 1) scad closes without warning
<stealth_>
J24k10, you will have to wait for one of the devs to help you with this topic.
<stealth_>
btw, the link has an extra space so it gives 404 page.
<J24k10>
https://imgur.com/a/hFimkvm is this better (links work fine for me there is something changed on irc clients)
<J24k10>
it is an edge case as i couldn't get this with normal gaps only with the angles of the voronoi and only when using the -delta offset and only if the gap is the rounding size (which shouldn't be used at all) but something seems to align points in a way that causing a crash
teepee has joined #openscad
<J24k10>
I think i had something similar with the SVGs (of a dragonfly wing https://makerworld.com/en/models/563928) but thought this may be related to some bad points on the svg
<pca006132>
it seems that createTriangulatedPolySetFromPolygon2d is somehow giving a corrupted triangulation
<J24k10>
haha just though why is my design now taking so long to render .. still had manifold deactivated - Ü
<J24k10>
pca006132 i have seen that offset can cause gaps and points be positioned wrong probably because of float precision calculations - so multiple offset can amplify this.
<J24k10>
if you set fn=36 (Rund(.5,fn=36) ) the crash did not happen.
lostapathy has joined #openscad
<pca006132>
still cannot fix the bug, but I think I found another issue
<pca006132>
at least in createTriangulatedPolySetFromPolygon2d, prev->info() should be polyset->vertices.size() + last_idx;
<pca006132>
otherwise the index will be wrong
<pca006132>
but I have no idea why cdt is giving invalid result at the end when all inputs are seemingly correct
<pca006132>
I added a check to discard the invalid face, and it somehow works? manifold is not reporting manifoldness error when I extrude that thing
<pca006132>
weird
mmu_man has joined #openscad
<J24k10>
sounds good - i wouldn't care about an error - but a crash is annoying ( and then i wish the backup file was top of the recent files)
<J24k10>
pca006132 sorry i have no environment to build/compile
<teepee>
no need, just download from the link
<pca006132>
you can probably try to wait for the CI
<J24k10>
teepee link? where? what?
<J24k10>
the nightly or something else?
<teepee>
J24k10: klick the pull-request link above
<teepee>
click the "checks" tab
<teepee>
make sure it has "CircleCI branch build" selected on the left side
<teepee>
click the link "openscad-mxe-64bit - Success"
<teepee>
click "Artifacts" tab
<J24k10>
teepee awesome thanks!
<J24k10>
pca006132 sorry still crashes without warning
hyperair has quit [Ping timeout: 260 seconds]
hyperair has joined #openscad
<pca006132>
is your code the same as the old one?
<pca006132>
at least your code works now on my machine
<guso78>
Suggest to Install msys2 and Debug with gdb. IT can exactly Spot the call Stack where Crash happened
<guso78>
.
<pca006132>
iirc the CI build does not have debug info
kintel has joined #openscad
<kintel>
Hi all - I'm back online! Will catch up over the next few days. Let me know if anything warrants my attention
<J24k10>
pca006132 ok this is weird - the code runs at the first preview but crashes on the second .. i used the module and then wrote the same offsets into code so it works without the module (tried to make a test case without the lib) and on works the other don't
kintel has quit [Client Quit]
<pca006132>
Hi kintel, the crash J24k10 found is probably due to some issue with the constrained delaunay triangulation thing
kintel has joined #openscad
<kintel>
Using CGAL's triangulator, or Manifold's?
<kintel>
i.e. with or without -DUSE_MANIFOLD_TRIANGULATOR=1 ?
<J24k10>
this is now gambling - i got 4× F5 until the crash
<pca006132>
without
<pca006132>
just CGAL's triangulator
<pca006132>
you can check my PR that kind of tries to fix it but doesn't fix it completely
<kintel>
Is there a way of reproducing the crash? The issue doesn't specify a test case
<pca006132>
hmm it is not exactly 5253, but I think they are related since they crash in the same part of the code, will add reproducer
<pca006132>
done
<pca006132>
J24k10 may have other cases, because the current code runs fine on my machine with the PR
<J24k10>
pca006132 have you tried preview F5 several times (without changing the code) - as 2024.08.14 works only sometimes
<pca006132>
ah, it takes quite a few times to crash it
<pca006132>
but yeah after spamming F5 it eventually crashed
<pca006132>
weird but also not very surprising because I am pretty sure my "fix" is not a proper fix
<pca006132>
I feel like this is some sort of memory corruption...
<kintel>
can reproduce with the UB test case
kintel has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
guso78k has joined #openscad
guso78k has quit [Client Quit]
lastrodamo has joined #openscad
kintel_ has joined #openscad
L29Ah has quit [Read error: Connection timed out]
kintel_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
kintel has joined #openscad
<kintel>
If I build with asan enabled, it triggers an undefined memory read also for non-crashing models
guerd871 has quit [Read error: Connection reset by peer]
<pca006132>
ahh
kintel has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<pca006132>
anyway, I think I got some idea about why the vertex info can be some non-sense
<pca006132>
it seems that CGAL constrained Delaunay triangulator inserts some vertex to represent special constraints when your edges overlap or something
<pca006132>
and that vertex is contained in the output triangles
kintel has joined #openscad
<pca006132>
and its info() is uninitialized... so it is garbage
<kintel>
Yeah, even for non-crashing variants of the UB model, I can see odd triangulations
<kintel>
..and it eventually crashes after a handful of re-renders
<kintel>
Anyway, we have a Manifold implementation for this too, not sure if it's ready for prime time yet
kintel has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<pca006132>
there are quite a lot of changes in manifold recently, and we are about to move to double precision
<pca006132>
I think one way to fix the issue would be to change the Info struct into something with default constructor? so vertices not added by us will get a -1 index for example, and we can just ignore faces with those vertices for now (until we figure out why they are there...)
kintel has joined #openscad
<kintel>
pca006132 yeah, at least move it from an uninitialized memory read into something that fails more consistently :)
<kintel>
Anyway, I'll try to look into it this week
<pca006132>
ok it seems to work on my machine again now, no matter how many F5 I did
<J24k10>
I tried to test with simple geometry but that didn't caused the crash - also not every voronoi pattern caused it ( or every $fs/ $fn ) - seems i hit the UB sweet spot
<pca006132>
hmm, the triangulation is still weird though
<pca006132>
updated, seems to fix the weird triangulation. those vertices coming from nowhere seems to be clones of the last vertex, so probably something related to line segments touching, similar to 3d 2-manifold issue?
J24k10 has joined #openscad
<pca006132>
so I just copy the last id to the vertex coming from nowhere, and the triangulation looks fine now with no error
<pca006132>
with that vertex ID patch, I don't think the -1 check is still needed
<J24k10>
ThX for fixing! very appreciated.
kintel has quit [Ping timeout: 258 seconds]
califax has quit [Remote host closed the connection]
califax has joined #openscad
snaked has joined #openscad
ferdna has joined #openscad
JoelJoel has joined #openscad
Joel has quit [Ping timeout: 252 seconds]
JoelJoel is now known as Joel
teepee_ has joined #openscad
Joel has quit [Max SendQ exceeded]
teepee has quit [Ping timeout: 260 seconds]
teepee_ is now known as teepee
Joel has joined #openscad
fancsali has joined #openscad
Virindi has quit [Read error: Connection reset by peer]
Virindi has joined #openscad
stealth_ has joined #openscad
lastrodamo has quit [Quit: Leaving]
misterfish has quit [Ping timeout: 248 seconds]
J24k10 has quit [Quit: Client closed]
J24k10 has joined #openscad
fancsali has quit [Quit: Client closed]
TheCoffeMaker has quit [Ping timeout: 252 seconds]
ferdna has quit [Quit: Leaving]
TheCoffeMaker has joined #openscad
TheCoffeMaker has quit [Read error: Connection reset by peer]