Compute All - When to use ?

[TrippyLighting]

I've been working on this model intensively (20+ hours) for the last 2-3 days and Fusion 360 has been fighting back like mad.

It is a 90MB file that takes 5 minutes to load and 5minutes to save.

Converting the LED arm into an STL file it takes 5 minutes for the dialog to show up (medium) and another 5 minutes to save to disk.

I made a few minor modifications to one piece  (added a few fillets) and decided for good measure to run "Compute All", something ive done occasionally, when I remembered.

This blew the model apart! The timeline looked perfectly normal before, now my timeline is plastered with yellow and red icons.

 So the question is: When do I need to run Compute All to see where things are going wrong ?

[keqingsong]

@TrippyLighting Ooph...that looks painful to repair, but nonetheless, it's repairable. My best practice experience is to do compute all at every major milestone of the design process. As soon as I have something designed to a state that I feel comfortable saving, I do a compute all to make sure that all geometry is being referred to properly and everything checks out. 

Note that playing the timeline back and forth does not actually solve for issues, it merely just "plays" the sequence of actions that took place. Computer All actually solves for all the relationships and references and makes sure that the design is legit. 

You're best bet right now is to use the timeline marker and go to the first red error (which looks like to be a combine) and troubleshoot why it is red. Invite me to your project if you need help. 

[TrippyLighting]

Hi @keqingsong thanks for the offer!

While the software sometimes drives me to the brink of madness, the people beind it are insanely great!

[jeff_strater · Accepted solution]

Hi @TrippyLighting,

This is a very good question.  I can only answer for my own workflows.  But, what I tend to do is:

• As long as I am just creating new features (that is, not going back and editing), I tend not to ever use Compute All.  Outside of bugs in Fusion logic, you should never have errors or warnings if you have not edited anything.  That said, as Keqing suggests, whenever I finish a major addition, I will tend to do a Compute All just as kind of a model check.  Then I save it.
• Once I start making changes, then I do this more often.  Some of this is experience (and, in my case, unfair knowledge of the internals 🙂 ).  For instance, if I just edit a sketch dimension, and the basic topology of the sketch has not changed, I never bother with Compute All.  But, if I start editing a sketch or early feature in some large-scale way (deleting some sketch curves and adding new ones), then I get nervous and will resort to Compute All.

As you use Fusion more and more, you will get a feel for what kind of changes are safe and what kind are more risky.  You also get a feel for what to do and what not to do when editing early features to avoid reference failures (I would guess that 90% of compute failures are "unresolved references").

The odd thing is that, in theory, Compute All should never be needed at all.  We try very hard to never let Fusion get into a "compute pending" state.  In fact, we debated about even including it at all.  But, we were using it internally to test parametrics, so we decided "if we get use out of it, customers will as well".  To be honest, if behavior like what you are seeing exists, I would consider that a bug.  There should really never be a case where an edit does not immediately reveal an error, but Compute All does.  If you see those, we need to debug those and fix them.  Once Keqing has your model, we can do a bit of investigating to see if we can figure it out, assuming that is OK with you.

Jeff

[TrippyLighting]

@jeff_strater @keqingsong I've invited you two to the project as that is the easiest way to share the file.

The file name is : Ring-O-Star V2

Be gentle with me 😉

I am quite aware that my workstyle is messy and chaotic and the actual product is still in an advancing prototype phase. I'll cean up when I am done 😉

What I mean with messy is that I am using quite a number of references, such as projected edges, sketch elements etc. across components.

[jeff_strater · Accepted solution]

Hi @TrippyLighting, thank you for inviting me to the design.

There is good news and bad news.  The good news is that I was able to repair the model.  I did a Save As into the same project, with the name "Ring-O-Star V2 (repaired)".  Compute All works fine on this model.

The bad news is that I don't understand what was causing the problems in the first place.  The problems are all very early in the model, with the first two Form Features.  If I do a Compute All, both of those fail to produce a valid solid body.  I don't know for sure why.  My "repair" consisted only of editing those Form Features, adding a new TSpline body (a block), then exiting the Form edit (which takes a LONG time).  Then, I went back and removed the block body.  I had to do this for both of those Form features.  But, now, everything appears OK - Compute All works on this model just fine.

I would like to add permission to download this model for our dev and QA team to look at to try to figure out why it doesn't work.  We would assure you that it would only be used for debugging purposes.

Thanks,

Jeff

[neljoshua]

I had the same question about when to use "Compute All", but my issue was the reverse; the model failed when I tried to "Get latest" until I performed a "Compute All".  Thanks @jeff_strater for the clarification.

I totally agree with @TrippyLighting regarding the superior quality of those developing and helping users.

[TrippyLighting]

@jeff_strater Thank you very much for looking into and takinge the time to repair. I'll look at it when I get home from work.

If it helps the software development process I am more than happy to contribute. You have my permission to download the file.

I sent the T-Spline based components - the LED arms and the Center Piece - off to the 3D printer on Sunday and hope to bring them with me in February and actually hope I can bring a complete assembled prototype, so maybe you can  observe these physical objects in the real world.

[TrippyLighting]

Now that I've re-read what you did to fix the model, I don't think that this "solution" would have occured to me at all.

Can you describe the thought process that led you to editing the T-Splines ?

Does your unfair advantage of knowing Fusion 360 internals have anythig to do with that ?

[jeff_strater]

More like "let's try a few random changes to see what happens".  The credit actually goes to a colleague who was looking over my shoulder at the time, trying to help debug.  He said "it doesn't look like the conversion from TSplines to solid happens in the Compute All - what would happen if we forced it by adding a new body to convert".  Did the trick.

Thanks for letting us look at your model.  It would be good to get to the bottom of this.

Jeff

[TrippyLighting]

Aha. Now the natural next question is how did your colleague see that the T-Spline conversion did not happen through the compute all.

What's the clue to look for ?

[jeff_strater]

It's hard to generalize a debugging process.  But, when I want to debug a broken model, I often roll back the timeline to the beginning and manually step through the features.  Sometimes one at a time, sometimes a group at a time.  Often, I'll only go back to just before the first error.  If it's my own design, I can usually recall what I was doing at each step.  For another person's design, it's harder (I've often though about how useful it would be to be able to put "comments" on timeline items, but then, like any commenting feature, it would take diligence to keep them accurate).  In this case, it was pretty obvious what was going on in the design, certainly in those first features.  So, I rolled back to just after the two first Form features.  I could edit the Form feature (though editing had some problems - I couldn't select TSpline faces, etc - that's one of the bugs that we are looking at).  If I exited Edit Form, the model looked OK, though.  So I tried a Compute All, and... no solid.  That's when I called my colleague over.  "Why can I edit the TSpline, and Exit works, but not Compute All?".  I think he just looked at the result and noticed that there was no solid body, when there was before the Compute All.  The only way that can happen is if the convert fails (but there was no error message, so that's weird), so he just figured that the conversion must not have been attempted at all.  We also noticed that  the Compute All was very fast.  That TSpline body is a monster, and it should take a long time to convert (which, in fact, it does), so that was another clue.  So, all the evidence pointed in that direction.

I'm looking forward to your visit in Feb. and seeing your cool stuff.

Jeff

[TrippyLighting]

Thanks again for the help!

I renamed the old file before the reapair by adding "(deprecated)" to the file name and removed the (repaired) from the new file. Just in case you want to look at it again as it still creating problems 😉

I simply wanted to remove the fillet from the PCB. I have not referenced any geometry off that fillet or PCB and Fusion 360 complains about numerous issues with yellow and red icons, and a host of warning messages on things that are not at all related to that fillet. Here is a littel screencast of the action.

The file has become pretty much un-workable ad I am wondering waht's the best path forward. The only thing that's left to do for this design revision is really the PCB shape. Then I'll export the PCB as a DXF and import it as an outline into Altium Circuit Maker (or maybe I use Eagle again for this one).

I tried to export (save copy as) the LED arm only but that resulted in a useless file with many errors and seeing how it's created I can see why "save copy as"  would not work.

The next thing I could try is to save-copy-as all of the components except of the three T-Spline based components and then re-assemble all of that in a new assembly with at least the T-Spline bodies as linked components. My thinking is that this should eliminate the long compute-all run times and make working with the file much faster.

Does that sound like a sensible approach ?

[TrippyLighting]

@jeff_strater

I really need a hand! Any attempts to recover even parts of this assembly lead to some dead end.

This time I simply wanted to add a little more thickness to the top surface of the Center Piece to be able to print it out of a different material.

But I get rebuild errors on entirely unrelated geometry.

[jeff_strater]

Hi @TrippyLighting, I will definitely take a look at it.  There is a bug in the current production build which is being fixed in today's update, that might be related, but I'll check out your model and see what I can find.

Jeff

[jeff_strater]

I'm seeing the original problem here, at the very least (those early TSplines are failing again).  I'm going to see if I can open it in the latest dev build, which will reflect what is going out in the update, in just a couple of hours.  Stay tuned.

Jeff

[TrippyLighting]

Thank you again for the help!

[jeff_strater]

So, good news and bad news...

In the dev build the problems still appear at first.  I spoke to the developer that made the fixes.  While the underlying bug has been fixed, the fix does not automatically repair the models.  So, here is what I believe the fix will be:

After tonight's update, open the model, and go through one more round of editing those 3 Form Features at the beginning of the timeline.  Do the trick like before to force a re-convert.  The safest seems to be just add a new body, although that will require another round trip to delete that body.  I'd roll back to just after those 3 Form Feature to keep the updates to a minimum.  So, what I did on the dev build (to a copy of your design) was to add a body to each of the 3 Form Features, exit, go back in, delete the extra body, then roll to the end, and do Compute All.  Then save.  What I've been told is at that point, after the update, your model will be fine.  I'd be happy to fix up your design again, after the update, just to spare you the pain.

I apologize for this.  I'm embarassed by this one, because it's in my area (modeling).  

Jeff

[TrippyLighting]

Hi Jeff, no need to be embarassed. If my model could help detect a bug then it's worth the pain.

I'll will try tomorrow evening as you suggested and see if I can repair the model.

[TrippyLighting]

It did not work.

I rolled the timeline back to the the "ground" after the third T-Spline.

I added three blobs to each of the T-Splines.

Then deleted the three blobs from the three T-Splines.

After that I rolled the timeline back to the end.

6 warnings, 12 errors.

I will not save the Ring-O-Star V2 file, so you can try your luck.