Phooey, my FreeCAD broke!

Ok, I see what you're talking about. I was confusing the 3.45mm bore and the 1.78. Nice that you got a developer to respond that quickly.
Sorry to have written in a confusing manner. When you are staring at something all day and you try to describe it, sometimes one leaves things out. Yes, that it was very good to have gotten the developer's eye. Hope it eventually fixes the problem.

Hard for me to figure out how to go from their git commit number to whatever is on github, don't understand the hash stuff, or more importantly how to know if the pull I have includes the fixes. I might have pulled from the wrong branch for all I know. The dev didn't say which branch, and I assumed it was main... Maybe not. Will have to ask.
 
The changes made by the developer did fix the seg fault, so that's great. It didn't fix the artifact problem, so I posted a separate thread in the Tech Forum area for FreeCAD. Maybe with a little luck that will get fixed.

Seems a bit weird that a silly user like me can make these bugs come out. It's not like I'm trying to break anything. I'm just inelegantly brute forcing my way through trying to get a model. I don't know the right way of doing it, I'm just trying anyway to do it.
 
So far, (if I ask the question in the right way,) the support on the FreeCAD forum has been very good. A developer confirmed the artifact and is packaging up a bug report for the library used by FreeCAD, called OCC (Open Cascade). It's a bug with hidden line removal. (My labeled artifacts!)

In any forum, the key to receiving a good answer is framing a good question that has the observation, what the results should be, and the actual results. Doing that succinctly is difficult, especially for the original poster, since they are often frustrated by the actual results. But we need to persevere, because without a well formed problem statement, it's hard to determine cause, or a solution. It is a lesson to be taken to heart.
 
Sorry to have written in a confusing manner. When you are staring at something all day and you try to describe it, sometimes one leaves things out. Yes, that it was very good to have gotten the developer's eye. Hope it eventually fixes the problem.
Well, I took it as a reading comprehension problem ;)

Seems a bit weird that a silly user like me can make these bugs come out. It's not like I'm trying to break anything. I'm just inelegantly brute forcing my way through trying to get a model. I don't know the right way of doing it, I'm just trying anyway to do it.
That's my approach to it. If there is a good comprehensive guide to using FreeCad, I have yet to find it. I was playing with the Arch (architecture) workbench recently. The wiki tutorial is from version 0.14, poor match to the current tools. I've tried several of the youtube videos but they are hard to use as reference material. But the alternative, using fusion360, isn't attractive either.
I don't think FreeCad has hit the point yet where Kicad is, enough community support to hit critical mass. Not that Kicad is perfect.
 
Well, I took it as a reading comprehension problem ;)


That's my approach to it. If there is a good comprehensive guide to using FreeCad, I have yet to find it. I was playing with the Arch (architecture) workbench recently. The wiki tutorial is from version 0.14, poor match to the current tools. I've tried several of the youtube videos but they are hard to use as reference material. But the alternative, using fusion360, isn't attractive either.
I don't think FreeCad has hit the point yet where Kicad is, enough community support to hit critical mass. Not that Kicad is perfect.
FreeCAD is ok for hacking around, and mostly good enough for my use. It shouldn't be confused with a commercial product though, it's not professional enough, and there are still quite a few issues with it. Excuse my little rant...

For one thing, only using a single core these days, well that's not how it should be done. Many computers, including my laptop, have access to hundreds, if not thousands of GPU cores to do work, (1920 GPU cores) never mind the 15 other idle CPU cores on my i7. Seems absurd to only use one single solitary core for computation. You and I both know that writing massively parallel code is very hard, but even some simple stuff could be done to make things a lot snappier. There has to be some bottlenecks that could be rewritten to make them faster by using parallelism.

We don't often use real helical threads in our drawing, why? Because we (FreeCAD) are stuck in the stone age using a single core... And lets not talk about the topological naming problem, or easy chamfering, or whatever. For hobbyist use FreeCAD is mostly ok. If I had a business, I wouldn't consider it at all, it just isn't mature enough to bet my business on.

This is not to say that other CAD packages are great, but they are quite a bit more full featured and mature. Most are a bit more expensive. Some are free now, or partially crippled, or will be crippled when times are tough and the companies need to bring up their margins. But generally, they are superior performing offerings, which is why businesses use them. FreeCAD is pretty good, but it doesn't perform at professional levels, at least in my limited perspective. Rant is over.

On the other hand, with FreeCAD, I can post on a forum, and get the personal attention of a developer. There is value to that. Still, the code is not quite finished.

When I see what others can do on HM and see their renderings, I am left drooling. I have no idea how to get FreeCAD to that level. Have to go reading a lot on rendering, because I know very little about it.

KiCAD has plenty of quirks. One of them is a pathological aversion to auto-routing. According to many there, only real men route. For crying out loud, I used rules based auto-routers professionally in 1985 for a first pass success. Yes it wasn't perfect, but using it was 3 times faster than hand routing estimates, even after fixing a few weirdly routed nets. It's like they are the only male bastion there... Got lambasted there for asking about auto-routing. Like I would enjoy hand routing, if I only got into it. Blechh, yes, it's problem solving, but I'd rather be doing something else, like using my completed board, three weeks earlier. Libraries, at least in KiCAD V6 are weird, don't know if that is much better in V7. Other than that, KiCAD is pretty good, been able to do most of what I need.
 
I suspect what really limits FreeCad is the Open Cascade dependency. I recall reading on some forum that Open Cascade is the issue that limits it to one thread. Open Cascade looks like an open source license of a company that was bought out (Matra). Given that both FreeCad and Open Cascade have been around for 20+ years, I don't expect to see rapid improvements.

If I ever get the VMC running and actually start using CNC, I may at that point have to re-evaluate which CAD package to use. What's your experience with FreeCAD driving your 3D printer?

Skeptical that the Kicad 7 database features for libraries will improve anything at the individual/hobbyist level.
 
FreeCAD works great for exporting to stl. Basically, just select the body or part and export to stl. Easy as pie. Haven't used their slicer, I use the Prusa Slicer.

Yes, I think FreeCAD's dependency on Open Cascade is what's keeping them back. Single core is pretty old fashioned. We were using GPU farms for whole vehicle EM simulation at mmW frequencies about six years ago. 30,000 GPU cores, now that's a little speed up. It made seemingly intractable models simulatable. Even if FreeCAD can't do GPU cores, it sure would make a big difference if they could use say half the CPU cores. The current single core implementation simply limits the model sizes and complexity to hobby size systems.
 
Seem to be good at breaking things. Attempted to add netgen as an alternative meshing tool for FEM. This turned into a cmake mess. As I got deeper into the pit, with more and more things breaking, I decided to flush it all down the drain and do a fresh build. Took a while, but I'm back in business, without netgen. Turns out netgen is broken, and has been for a while.

gmesh seems to work in FEM Workbench, at least if you have a decent model.
 
Back
Top