ali12341 has quit [Remote host closed the connection]
ali1234 has joined #openscad
guso78 has quit [Read error: Connection reset by peer]
guso78 has joined #openscad
Junxter has joined #openscad
J25k92 has quit [Quit: Client closed]
J25k92 has joined #openscad
L29Ah has joined #openscad
mtm has quit [Ping timeout: 252 seconds]
mtm has joined #openscad
mmu_man has joined #openscad
<J25k92>
teepee using color instead of base material for 3mf export .. crashes openscad
<J25k92>
2025.01.24
<J25k92>
and every object got the color of the first
Non-ICE has quit [Quit: Screw you guys, im going home!]
drkow has joined #openscad
teepee_ has joined #openscad
drfff has quit [Ping timeout: 252 seconds]
teepee has quit [Ping timeout: 264 seconds]
teepee_ is now known as teepee
Non-ICE has joined #openscad
<J25k92>
teepee there is an issue when exporting color instead base material - oSCAD crashes on WIN (2025.01.24)
buZz has quit [Remote host closed the connection]
knielsen has quit [Ping timeout: 265 seconds]
knielsen has joined #openscad
kintel has joined #openscad
<kintel>
Oh, I guess we didn't yet add tests for export options. Dealing with options is a bit of a gap in the test framework, should find a better way do deal with those.
mmu_man has quit [Read error: Connection reset by peer]
Guest71 has joined #openscad
mmu_man has joined #openscad
Guest49 has joined #openscad
<Guest49>
heyy
<Guest49>
anyone here??
<teepee>
yes
<teepee>
J25k92: I'll have a look later, manual tests on linux seemed ok, but I had some issues with different builds earlier too
<juri_>
indeed.
kintel has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<teepee>
if the files are there and can be opened as usual, that's already the full regression test
snaked has joined #openscad
<teepee>
I added the option to store user defined examples in the home directory, like for code snippets
<J25k92>
the user examples are not there but i can open the files (tested 1)
<teepee>
but that needed some changes in the handling of the existing example stuff too, and reading folders and files is always somthing that can do surprising stuff on different platforms
<teepee>
for the user files you would need to recreate the user examples folder as shown in the library info window and add a subfolder + a fle inside
<J25k92>
(the examples could be more colorfull)
<teepee>
or file ;-)
<teepee>
what was the 3mf crash, did the model have colors?
<teepee>
so DeepSeek is useless for OpenSCAD too :-) I asked for a fume extractor using a 40x40mm fan and it basically created 2 cubes :D
<J25k92>
oh that is nice .. the folder under the user documents "examples" are there
<J25k92>
teepee the crash was that when choosing "colors" instead of base material it closed without warning after the save file dialog .. a corrupt file was created
<J25k92>
and when using base colors it works but i had two objects (different colors) and both got the color of the first
<gbruno>
[github] stefano2734 opened issue #5625 (New release for moving target openscad development since 2021 please in 2025, python 3.8 is eol) https://github.com/openscad/openscad/issues/5625
<teepee>
hmm, no, but having a similar feature for backups would be nice
<teepee>
I suppose that would be more like a table with information when that file was saved and so on
kintel has joined #openscad
<kintel>
In terms of backups, it might be possible to determine that OpenSCAD terminated uncleanly, and use that to offer loading a backup. As with all GUI code, the devil is in the details. I think someone opened a ticket requesting this a while ago.
kintel has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
mmu_man has quit [Ping timeout: 265 seconds]
<church__>
regarding GUI, would be nice to see distance & angle measure to be changed to "toggle switches", like on/off axis, or choice perspective or orthogonal projection. PITA that one cannot switch those measurement tools off, once used, unless reloading model or restarting app
mmu_man has joined #openscad
<church__>
again regarding GUI, would be nice to see axis always shine through model. and measurement numbers turn around axis to stay perpendicular to vector of view, including small notches of units. This way, even if one is from side of axis, and even if axis goes behind/in model, one can still see them & still see numbers
<church__>
(or at least those small unit notch intervals, without rotating model for them to appear)
<InPhase>
kintel: One devil in the details... If the user clicks yes to restore it, the editor has to be unhidden. :)
<teepee>
yep, that would totally make sense
<InPhase>
It would take a while to notice that bug if it's missed, because the situation doesn't show up normally with external editor use and automatic reload and preview.
<InPhase>
But then once in a while someone would trigger it.
<InPhase>
You have to edit, not save, hide the editor, render, and crash.
<teepee>
in other news, the new hype deepseek has very kind words about openscad but is still just horribly useless when asked for a model ;-)
<InPhase>
When AI takes everyone's jobs, we'll just have to become OpenSCAD model makers, the last refuge from the AI jobpocalypse.
<teepee>
so far that seems a pretty safe niche, but also one no one cares about?
<InPhase>
Which is probably what keeps it that way. :)
<church__>
or just like ms word/excel in past generated horrible webpages, AI might generate openscad models full of minkowski, non manifold and so on :)
<teepee>
so far it explains very nice features, but generates pretty much any model by combining a couple of random cubes
<teepee>
the source is nicely documented though and it did actually get the syntax right on the first try
<InPhase>
teepee: If we wanted the problem solved though, we just need to make an OpenSCAD model benchmark, and convince the model evaluators to start using it in their pile of metrics as a visuospatial reasoning metric.
<InPhase>
teepee: Then the people making the models will get mad that they're getting a low score on this, and tune the models to this benchmark.
<church__>
until one gets to bits that are hard or PITA to do in. like filleting/chamfering complex shapes, geometry to orient or attach part to other part, lack of concave hull and so on .. then it gets 10x more time consuming and often reason to switch for less patient
<church__>
can one still do that in openscad? sure. but it deviates far from combining random cubes, with 10x more spaghetti code added on top
<InPhase>
church__: Often the solutions remain very elegant, but require different modalities of reasoning. And these are definitely in scope of what current gen AI models do not even come close to excelling at currently.
<teepee>
hmm, raspi5 is not good enough for the 14b model, I wonder if the 16GB one would make a difference