Regression: Inconsistent x-ray wireframe edge thickness on symmetrical mesh #119527

New Issue

Andrea-Lanaro · 2024-03-15T17:37:24+01:00

Andrea-Lanaro commented

2024-03-15 17:37:24 +01:00

System Information
Operating system: Windows-10-10.0.17763-SP0
Graphics card: NVIDIA GeForce RTX 3060/PCIe/SSE2

Blender Version
Broken:

4.1 alpha 44ac2e8a8d
4.1 beta c1f55288e9
4.1 rc
Worked:
4.0.2
3.6.9 LTS

Short description of error
Viewport x-ray wireframe of symmetrical object have inconsistent edge thickness when view from its axis of symmetry.

In 3.6 and 4.0, these edges overlap without aliasing and stay consistent at different zoom level.

version
4.0
4.1

Exact steps for others to reproduce the error
Add primitive with expected symmetry. View in wireframe, x-ray, orthographic.

**System Information** Operating system: Windows-10-10.0.17763-SP0 Graphics card: NVIDIA GeForce RTX 3060/PCIe/SSE2 **Blender Version** Broken: - 4.1 alpha 44ac2e8a8d24 - 4.1 beta c1f55288e9cc - 4.1 rc Worked: - 4.0.2 - 3.6.9 LTS **Short description of error** Viewport x-ray wireframe of symmetrical object have inconsistent edge thickness when view from its axis of symmetry. In 3.6 and 4.0, these edges overlap without aliasing and stay consistent at different zoom level. | version || |--|--| | 4.0 | ![](attachments/5fa0440d-5d07-4032-bee9-3552c4e7b41d) | | 4.1 | ![](attachments/eed2de0d-8239-46ff-97fa-b73d3bf91781) | **Exact steps for others to reproduce the error** Add primitive with expected symmetry. View in wireframe, x-ray, orthographic.

orthowireframe-402.png

329 KiB

orthowireframe-410.png

348 KiB

wiretest.blend

1.0 MiB

Andrea-Lanaro added the

labels 2024-03-15 17:37:25 +01:00

Germano Cavalcante commented

2024-03-15 18:53:41 +01:00

I tested different versions, but in my case, I can't notice any difference, it must depend on some monitor or GPU configuration.

**System Information**
Operating system: Windows-10-10.0.22621-SP0 64 Bits
Graphics card: NVIDIA GeForce RTX 3060 Laptop GPU/PCIe/SSE2 NVIDIA Corporation 4.6.0 NVIDIA 546.01

I tested different versions, but in my case, I can't notice any difference, it must depend on some monitor or GPU configuration. ``` **System Information** Operating system: Windows-10-10.0.22621-SP0 64 Bits Graphics card: NVIDIA GeForce RTX 3060 Laptop GPU/PCIe/SSE2 NVIDIA Corporation 4.6.0 NVIDIA 546.01 ```

Richard Antalik commented

2024-03-15 20:51:42 +01:00

Also can't reproduce (Linux, RTX 3060)
Monitor would probably affect UI scale, and even with scaling value of 2 I can't reproduce. But it would be useful to see if any DPI scaling has any influence on this. @Andrea-Lanaro can you check that?

Also can't reproduce (Linux, RTX 3060) Monitor would probably affect UI scale, and even with scaling value of 2 I can't reproduce. But it would be useful to see if any DPI scaling has any influence on this. @Andrea-Lanaro can you check that?

Richard Antalik added

Status

Needs Information from User

and removed

Status

Needs Triage

labels 2024-03-15 20:51:52 +01:00

Andrea-Lanaro commented

2024-03-15 21:27:04 +01:00

Same result on two of my display, Dell U2410 and Cintiq 13 HD.
Both are on their native resolution, 1920x1200 and 1920x1080. No scaling applied on OS level.

Tried with OS level scaling on, still same result.

Updating GPU driver have no effect. Please let me know if there is anything else I can try.

Lines in edit mode are identical in 4.0 and 4.1

Same result on two of my display, Dell U2410 and Cintiq 13 HD. Both are on their native resolution, 1920x1200 and 1920x1080. No scaling applied on OS level. Tried with OS level scaling on, still same result. Updating GPU driver have no effect. Please let me know if there is anything else I can try. Lines in edit mode are identical in 4.0 and 4.1

Richard Antalik commented

2024-03-15 21:42:44 +01:00

Eh sorry, I just checked 4.2. Checking 4.1 just in case, and indeed it has this bug.

It is interesting, that is not an issue in 4.2...

Eh sorry, I just checked 4.2. Checking 4.1 just in case, and indeed it has this bug. It is interesting, that is not an issue in 4.2...

❤️ 1

Richard Antalik added

and removed

Needs Information from User

labels 2024-03-15 21:43:02 +01:00

Richard Antalik changed title from ~~Inconsistent x-ray wireframe edge thickness on symmetrical mesh~~ to Regression: Inconsistent x-ray wireframe edge thickness on symmetrical mesh

2024-03-15 21:43:20 +01:00

Richard Antalik commented

2024-03-16 18:14:50 +01:00

Caused by 14500953ed
CC @xavierh

Caused by 14500953ed5a6ecbdc5842c9f86c557b1f6af5e1 CC @xavierh

Xavier Hallade commented

2024-03-18 09:41:25 +01:00

14500953ed is present in 4.2 that doesn't exhibit the issue.

@fclem any idea what in 4.2 fixes this scenario ?
If not, I suppose we can revert 14500953ed in 4.1 specifically.

14500953ed is present in 4.2 that doesn't exhibit the issue. @fclem any idea what in 4.2 fixes this scenario ? If not, I suppose we can revert 14500953ed in 4.1 specifically.

Jeroen Bakker self-assigned this 2024-03-18 14:57:36 +01:00

Jeroen Bakker added this to the EEVEE & Viewport project 2024-03-18 14:57:58 +01:00

Jeroen Bakker commented

2024-03-19 09:30:32 +01:00

I looked at the differences and it seems like a normal buffer is attached in Blender 4.1, but not in Blender 4.2. An attached normal buffer will change the wireframe drawing at https://projects.blender.org/blender/blender/src/branch/main/source/blender/draw/engines/overlay/shaders/overlay_wireframe_vert.glsl#L79

Simulating the shader to use the other part makes the render work correctly so the incorrect normal buffer is the cause that the render is failing. Unsure what changed its behavior as bisecting is hard to do with git-lfs...

I believe #116902 might have fixed it when the normals and positions have been separated. This isn't something I want to add to the release last minute. Adding a work-around (adding a fix for 4.1 solely) also introduces a risk as wireframe rendering is used all the place and we would introduce artifacts on other places.

So if reverting the performance improvements in 4.1 would solve the issue, we should do that. Looking at the PR this would already been an issue for several years. @fclem what do you think?

I looked at the differences and it seems like a normal buffer is attached in Blender 4.1, but not in Blender 4.2. An attached normal buffer will change the wireframe drawing at https://projects.blender.org/blender/blender/src/branch/main/source/blender/draw/engines/overlay/shaders/overlay_wireframe_vert.glsl#L79 Simulating the shader to use the other part makes the render work correctly so the incorrect normal buffer is the cause that the render is failing. Unsure what changed its behavior as bisecting is hard to do with git-lfs... I believe https://projects.blender.org/blender/blender/pulls/116902 might have fixed it when the normals and positions have been separated. This isn't something I want to add to the release last minute. Adding a work-around (adding a fix for 4.1 solely) also introduces a risk as wireframe rendering is used all the place and we would introduce artifacts on other places. So if reverting the performance improvements in 4.1 would solve the issue, we should do that. Looking at the PR this would already been an issue for several years. @fclem what do you think?

Jeroen Bakker commented

2024-03-19 13:24:16 +01:00

Just checked with Clement. I will revert the commit in 4.1, but keep it in 4.2.

The issue is that the optimization introduced a bug, but was hidden in 4.2 when the normal and position vbo where separated.

Just checked with Clement. I will revert the commit in 4.1, but keep it in 4.2. The issue is that the optimization introduced a bug, but was hidden in 4.2 when the normal and position vbo where separated.

Jeroen Bakker referenced a pull request that will close this issue

2024-03-19 13:33:18 +01:00

Fix #119527: Aliased Wireframe In XRay #119656

Jeroen Bakker referenced this issue from a commit

2024-03-19 14:23:52 +01:00

Fix #119527: Aliased Wireframe In XRay

Jeroen Bakker closed this issue

2024-03-19 14:23:56 +01:00

Blender Bot added

and removed

labels 2024-03-19 14:23:59 +01:00

Jeroen Bakker referenced this issue from a commit

2024-03-19 14:31:45 +01:00

Fix #119527: Aliased Wireframe In XRay

Jeroen Bakker referenced this issue from a commit

2024-03-19 14:31:45 +01:00

Revert "Fix #119527: Aliased Wireframe In XRay"

Jeroen Bakker referenced this issue from a commit

2024-03-19 14:37:07 +01:00

Revert "Fix #119527: Aliased Wireframe In XRay"

Iliya Katushenock commented

2024-03-19 15:26:38 +01:00

Not sure about things in 4.1 and 4.2 @Jeroen-Bakker , report was about 4.0 and 4.1 or there is some other details?

Jeroen Bakker commented

2024-03-19 15:33:00 +01:00

Fix for this issue only landed in 4.1, but as 4.1 must always be merged into 4.2 it was manually reverted again in 4.2. So this fix only landed in 4.1. If anything went wrong, please let me know ASAP

Fix for this issue only landed in 4.1, but as 4.1 **must** always be merged into 4.2 it was manually reverted again in 4.2. So this fix only landed in 4.1. If anything went wrong, please let me know ASAP

Iliya Katushenock commented

2024-03-19 15:35:26 +01:00

I means, regression was between 4.0 and 4.1, but you fix something which is occur between 4.1 and 4.2

Jeroen Bakker commented

2024-03-22 08:22:09 +01:00

@mod_moder please test, and report back if something isn't working. Git logs can be confusing in our workflow to find out the current state and what actually changed and therefore doing a theoretically report might not be correct.

👍 1

Andrea-Lanaro commented

2024-03-30 12:53:25 +01:00

@Jeroen-Bakker Looks fine in 4.1 stable. I found something in 4.2 d7c718dfd4.
The subdivision modifier brings back aliased edge.

This only happen if GPU subdivision is enabled.

@Jeroen-Bakker Looks fine in 4.1 stable. I found something in 4.2 d7c718dfd467. The subdivision modifier brings back aliased edge. This only happen if GPU subdivision is enabled.

Andrea-Lanaro referenced this issue

2024-04-03 15:01:15 +02:00

GPU Subdivision modifier produce aliased wireframe in xray #120208

Germano Cavalcante referenced this issue

2024-04-03 22:05:31 +02:00

GPU Subdivision modifier produce aliased wireframe in xray #120208

Jeroen Bakker referenced this issue

2024-04-05 08:07:33 +02:00

Revert "Overlay: Wireframe: avoid using custom depth bias with xray enabled" #120285

Jeroen Bakker referenced this issue from a commit

2024-04-05 08:08:20 +02:00

Revert "Overlay: Wireframe: avoid using custom depth bias with xray enabled"