Cura: 4.0.0 BETA 2: Thin Walls / Gap filling broken

[BagelOrb]

Application Version
Cura 4.0.0 Beta 2

Platform
Linux

Printer
UM3

Steps to Reproduce
Enable Print Thin Walls
Load model
(if loaded from image file) Set thickness to 0.3 to 2.0mm

UM3_heman hueman 3__zw.3mf.rename.zip

Actual Results
Thin walls are only jagged lines

4.0.0-beta2 from attached curaproject

Expected results
Jagged lines are combined into smooth lines

4.0.0-Beta

Additional Information
I have tested and verified that this was not a problem in Cura 3.4, Cura 3.6, or even Cura 4.0.0 BETA 1

[smartavionics]

Hi Tim, the thin wall (and wall gap filling) in 4.0 under-performs (being polite here). My alternative implementation can do much better (YMMV). If you do try mine and it doesn't work OK, please let me know. Thanks.
https://www.dropbox.com/sh/s43vqzmi4d2bqe2/AAADdYdSu9iwcKa0Knqgurm4a?dl=0

[BagelOrb]

I am saying that thin walls in 4.0-beta works as normal, but it is 100% broken in 4.0-beta2

[smartavionics]

I'm still saying that the thin wall (and gap filling support) in 4.0 under-performs.

[BagelOrb]

I am planning on fixing the thin walls within the next year integrally.
I hope my solution will outperform even yours.

I remember we already talked about your gap filling algorithms somewhere. Do you remember which thread?

[BagelOrb]

Can you give a quick recap of the general idea of your implementation of gap filling?

[smartavionics]

Sorry, I can't remember the issue where we discussed my implementation.

I don't know if I will ever submit it as a PR as I know that in some situations it will not produce nice output and also I think it's quite slow. but for most simple cases it works well.

So, for both thin walls and gap filling, I estimate the points that define the centre line of the region that needs to be filled and I also estimate how wide the region is at those points. I also calculate the area that is drawn so that when the lines are output it avoids drawing over the same region twice.

[BagelOrb]

This is what it should look like - a lithophane of my ultimaker forum logo.

[smartavionics]

I do think that the transitions from fat to thin walls are very tricky to print nicely.

[BagelOrb]

Mark Burtons version:

4.0.0-beta:

[smartavionics]

So why did my comments get zapped?

[BagelOrb]

Just cleaning up conversation which is irrelevant to this issue.
My misuse of your version of Cura is to no use for anybody.

[smartavionics]

OK, just worried that it was the thought police erasing my history!

[Ghostkeeper]

~Maybe caused by this or this?~ We should do a bisect.

Edit: Neither of those commits are in 4.0.

[BagelOrb]

The first one is unlikely the culprit. The second one is a more likely culprit.

[smartavionics]

It will definitely be related to the changes they made to MergeInfillLines as that is what is meant to turn the infill lines into beautiful smooth curves.

[BagelOrb]

So perhaps it was this one: https://github.com/Ultimaker/CuraEngine/commit/ca9e4ae581658a33a7265aee12e2f7d5f26eea63#diff-cd4c69d037508b392e5eff0749d08c45

[Ghostkeeper]

That commit was made before the first beta was released though.

[BagelOrb]

Ok well, then we should just do a bisect!

[Ghostkeeper]

One of our QA engineers is doing that right now, using daily builds. Problem is though: The bug reproduces on January 8th, but not on December 18th (the date of the 1st beta). But there have been no commits on CuraEngine in the 4.0 branch in that period.

[smartavionics]

Note that the hack I inserted to disable the path optimization when gyroid infill was used wasn't removed until 23rd Jan.

[smartavionics]

Although, Tim's profile isn't using gyroid infill anyway.

[BagelOrb]

Maybe the merge was in that period, while the commit was from earlier ;)

[BagelOrb]

Or maybe a magic evil ghost creeped into the build system.
We all know you tend to keep those around ;)

[Ghostkeeper]

We found it. It's this one: https://github.com/Ultimaker/CuraEngine/commit/6523774aa453e9dd2476991a671242fb6ba75182

That one was merged on the 18th (the day of the release of the first beta) but after we built the beta release, so we were confused by that.

We also discussed it (with the history in CURA-5776). I actually also found this issue a while ago and raised a ticket because there are certain thicknesses of parallel walls where this happens. We can't control all cases using the current method. We have two tickets open about this:

• One to split up the cases of thin walls from the cases of surfaces ending in a corner, a longer-term fix.
• One to tune the current algorithm better so that it at least works correctly for all parallel walls (like this case).

Neither have made it into 4.0 though.

[smartavionics]

https://github.com/Ultimaker/CuraEngine/commit/6523774aa453e9dd2476991a671242fb6ba75182 contains changes in lots of different files but none are in MergeInfilLines so which of those changes is the culprit?

[BagelOrb]

That's a merge commit! Are you saying the merge went wrong or that a bad commit was merged?

Anyway I think the least we should do is to revert the changes which were the culprit, because the intended fix was less bad than the currently broken feature.

[BagelOrb]

I skipped through the merge commit changes, but couldn't find the mistake...

[smartavionics]

But you can't revert that merge commit as it would trash loads of useful changes.

[BagelOrb]

@smartavionics Nice winter design on the top banner, btw ;)

[nallath]

The commit that caused the behavior was merged since the merge commit. Reverting it would fix this issue but it would re-introduce others.

[BagelOrb]

Then revert it and fix those others.

[Ghostkeeper]

I'd rather work in the current direction than to revert the 20 commits and re-do essentially the same fixes. As SmartAvionics said, it would throw quite a lot of useful changes away.

We already have three open issues to fix this: The two I mentioned before in this thread and Smartavionics' version. We intend to combine the three. That should be sufficient.

[BagelOrb]

Oh that's even better.

As long as it gets fixed before the 4.0 release.

Get Outlook for Androidhttps://aka.ms/ghei36

---

From: Ghostkeeper [e-mail]
Sent: Monday, March 18, 2019 5:08:15 PM
To: Ultimaker/Cura
Cc: Tim Kuipers; Author
Subject: Re: [Ultimaker/Cura] 4.0.0 BETA 2: Thin Walls suddenly broken (#5445)

I'd rather work in the current direction than to revert the 20 commits and re-do essentially the same fixes. As SmartAvionics said, it would throw quite a lot of useful changes away.

We already have three open issues to fix this: The two I mentioned beforehttps://github.com/Ultimaker/Cura/issues/5445#issuecomment-473919468 in this thread and Smartavionics' version. We intend to combine the three. That should be sufficient.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHubhttps://github.com/Ultimaker/Cura/issues/5445#issuecomment-473980047, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AIe9EYOH_6OpiTKWVLUSdOM7rtgYHMpZks5vX7nvgaJpZM4b5Wen.

[Ghostkeeper]

We will not be able to fix this before the 4.0 stable release (i.e. in about an hour).

[BagelOrb]

I think it is not allowable to have this bug in a stable release.
I'm not putting any effort in trying to delay an unstoppable train, though.

D:'

---

From: Ghostkeeper [e-mail]
Sent: Tuesday, March 19, 2019 13:53
To: Ultimaker/Cura
Cc: Tim Kuipers; Author
Subject: Re: [Ultimaker/Cura] 4.0.0 BETA 2: Thin Walls suddenly broken (#5445)

We will not be able to fix this before the 4.0 stable release (i.e. in about an hour).

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHubhttps://github.com/Ultimaker/Cura/issues/5445#issuecomment-474350077, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AIe9EZaZVHysLaMWx5c9kpbgFgX4z-C5ks5vYN3UgaJpZM4b5Wen.

[Ghostkeeper]

Yeah, sorry. If it makes you feel any better, the problem was worse in version 3.6 :sweat_smile: It occurred in different cases, but more frequently.

[BagelOrb]

But, but, ... I've tested it to be working in 3.6.

It is 100% broken in 4.0.0-beta2, how could it be worse?
The gap filling now always consists of ragged lines - they are never combined into smooth lines.

[BagelOrb]

Take a better look at my screenshots.

[BagelOrb]

I thought my test case was pretty representative for most cases where people have models with thin walls, but maybe you think my case is an edge-case..?

[BagelOrb]

I discussed it in person and we found that the specific case I present here is indeed not that representative for all thin walls / gap filling cases.

Now I don't consider this issue any different from the general issue that gap filling and thin walls are generally problematic.

[Ghostkeeper]

We've found that disabling the Fill Small Gaps setting causes the lines from Thin Walls to be joined properly. That may be a workaround for some (but filling small gaps is necessary for this lithophane anyway).

[LipuFei]

Fixed in https://github.com/Ultimaker/CuraEngine/pull/1105