Blender Render Texture and Material Views Show Large Wireframes when viewing OpenSubDiv Objects #46706

New Issue

Dennis Brown · 2015-11-05T22:11:45+01:00

Dennis Brown commented

2015-11-05 22:11:45 +01:00

System Information
Windows 10 Pro x64

Blender Version
Broken: 2.76 48f7dd6

When viewing an object using OpenSubDiv with Texture or Material mode in Blender Internal, the object is shown as if it was using a thick wireframe modifier.

Exact steps for others to reproduce the error
Setup: Create an object of any sort, add a subsurf with at least one subdivision. Enable OpenSubDiv.

Switch to Material View Mode, and observe the slightly larger wireframe view
Enable Texture mode, and notice that the object shows correctly
Change the Material Shading mode to GLSL, and the Texture Mode will now show a wireframe just like Material Mode
Disable OpenSubDiv at any point to allow the material to show again.

It seems that Material mode does this with both Multitexture, and GLSL, but Texture mode is only effected under GLSL.

Here's an example with a Cube subdivided twice:

**System Information** Windows 10 Pro x64 **Blender Version** Broken: 2.76 48f7dd6 When viewing an object using OpenSubDiv with Texture or Material mode in Blender Internal, the object is shown as if it was using a thick wireframe modifier. **Exact steps for others to reproduce the error** Setup: Create an object of any sort, add a subsurf with at least one subdivision. Enable OpenSubDiv. 1. Switch to Material View Mode, and observe the slightly larger wireframe view 2. Enable Texture mode, and notice that the object shows correctly 3. Change the Material Shading mode to GLSL, and the Texture Mode will now show a wireframe just like Material Mode 4. Disable OpenSubDiv at any point to allow the material to show again. It seems that Material mode does this with both Multitexture, and GLSL, but Texture mode is only effected under GLSL. Here's an example with a Cube subdivided twice: ![opensubdiv.jpg](https://archive.blender.org/developer/F252858/opensubdiv.jpg)

Dennis Brown commented

2015-11-05 22:11:45 +01:00

Changed status to: 'Open'

Dennis Brown commented

2015-11-05 22:11:45 +01:00

Added subscriber: @DBrown

Sergey Sharybin commented

2015-11-06 13:43:38 +01:00

Added subscriber: @Sergey

Sergey Sharybin commented

2015-11-06 13:43:39 +01:00

I can not reproduce the issue with the given steps. please:

Make sure your video driver is totally up to date
Provide information about your graphics card
Attach system-info.txt generated by Help -> System Info
Attach simple .blend file with small instructions of reproducing the issue from that .blend

I can not reproduce the issue with the given steps. please: - Make sure your video driver is totally up to date - Provide information about your graphics card - Attach system-info.txt generated by `Help -> System Info` - Attach simple .blend file with small instructions of reproducing the issue from that .blend

Dennis Brown commented

2015-11-06 19:01:15 +01:00

My sincere apologies:

CPU: AMD FX 8350
CPU Chipset: 990fx
Video Card: XFX AMD R9 390
Catalyst Version: 15.7.1 (Latest stable)

Tested latest and daily (same results):

2.76b #f337fea
2.76.2 #4b316e7

*I installed the AMD Catalyst 15.11 beta drivers, which also gave the same results.

I tried OpenSubDIv in both CPU, and GLSL Transform Feedback, and both gave the same results (GLSL Compute not available).

I'm sure it doesn't make any difference, but when deselected, it shows as bands (screenshots attached).

Here is the blend file, putting you exactly where I was in the screenshots:

osd_test.blend

and here is my system-info:

system-info.txt

My sincere apologies: CPU: AMD FX 8350 CPU Chipset: 990fx Video Card: XFX AMD R9 390 Catalyst Version: 15.7.1 (Latest stable) Tested latest and daily (same results): 2.76b #f337fea 2.76.2 #4b316e7 *I installed the AMD Catalyst 15.11 beta drivers, which also gave the same results. I tried OpenSubDIv in both CPU, and GLSL Transform Feedback, and both gave the same results (GLSL Compute not available). I'm sure it doesn't make any difference, but when deselected, it shows as bands (screenshots attached). ![opensubdiv_deselect.jpg](https://archive.blender.org/developer/F253547/opensubdiv_deselect.jpg) Here is the blend file, putting you exactly where I was in the screenshots: [osd_test.blend](https://archive.blender.org/developer/F253553/osd_test.blend) and here is my system-info: [system-info.txt](https://archive.blender.org/developer/F253552/system-info.txt)

Bastien Montagne commented

2015-11-13 16:31:27 +01:00

Added subscriber: @mont29

Bastien Montagne commented

2015-11-13 16:31:27 +01:00

Thanks for the info. For what is worth, cannot reproduce this here (nvidia 850m, debian64 testing). Probably something related to that GPU/driver?

Dennis Brown commented

2015-11-13 21:14:35 +01:00

Somehow this didn't submit earlier this week, but I have to agree it has to be a GPU issue, as I didn't have these issues with the nVidia 960 I replaced with the 390.

Here are the errors given when this happens:

GPUShader: compile error:
Geometry shader failed to compile with the following errors:
WARNING: 3:4: deprecated130(#55) "varing" is deprecated since GLSL1.3. We suggest usage of "in/out"
WARNING: 3:5: deprecated130(#55) "varing" is deprecated since GLSL1.3. We suggest usage of "in/out"
WARNING: 3:7: warning(#312) interface block with instance is supported in GLSL 1.5
WARNING: 3:30: warning(#312) interface block with instance is supported in GLSL 1.5
ERROR: 3:49: error(#202) No matching overloaded function found: texelFetch
ERROR: 3:49: error(#202) No matching overloaded function found: texelFetch
ERROR: 3:70: error(#202) No matching overloaded function found: texelFetch
ERROR: 3:70: error(#202) No matching overloaded function found: texelFetch
ERROR: error(#273) 4 compilation errors.  No code generated

Somehow this didn't submit earlier this week, but I have to agree it has to be a GPU issue, as I didn't have these issues with the nVidia 960 I replaced with the 390. Here are the errors given when this happens: ``` GPUShader: compile error: Geometry shader failed to compile with the following errors: WARNING: 3:4: deprecated130(#55) "varing" is deprecated since GLSL1.3. We suggest usage of "in/out" WARNING: 3:5: deprecated130(#55) "varing" is deprecated since GLSL1.3. We suggest usage of "in/out" WARNING: 3:7: warning(#312) interface block with instance is supported in GLSL 1.5 WARNING: 3:30: warning(#312) interface block with instance is supported in GLSL 1.5 ERROR: 3:49: error(#202) No matching overloaded function found: texelFetch ERROR: 3:49: error(#202) No matching overloaded function found: texelFetch ERROR: 3:70: error(#202) No matching overloaded function found: texelFetch ERROR: 3:70: error(#202) No matching overloaded function found: texelFetch ERROR: error(#273) 4 compilation errors. No code generated ```

Mike Erwin commented

2015-11-14 22:30:41 +01:00

Added subscriber: @MikeErwin

Mike Erwin commented

2015-11-14 22:30:41 +01:00

Same problem here. The "thick wireframe" is the standard selection outline. Usually you'd see just the silhouette, but when something goes wrong...

One of our shaders fails to compile when switching to material display mode. Also matcaps in solid display mode.

Intel driver's GLSL compiler on Windows 10 complains:

GPUShader: compile error:
ERROR: 3:8: 'varying' : cannot be a struct
ERROR: 3:27: 'varposition' : undeclared identifier
ERROR: 3:27: 'assign' : cannot convert from '3-component vector of float' to 'float'
ERROR: 3:28: 'varnormal' : undeclared identifier
ERROR: 3:28: 'assign' : cannot convert from '3-component vector of float' to 'float'
ERROR: 3:42: 'outpt' : undeclared identifier
ERROR: 3:42: 'v' : field selection requires structure, vector, or matrix on left hand side
ERROR: 3:42: 'position' : field selection requires structure, vector, or matrix on left hand side
ERROR: 3:42: 'assign' : cannot convert from '4-component vector of float' to 'float'
ERROR: 3:43: 'v' : field selection requires structure, vector, or matrix on left hand side
ERROR: 3:43: 'normal' : field selection requires structure, vector, or matrix on left hand side

According to grep, "varposition" is found in gpu_shader_[geometry|vertex|vertex_world].glsl

Same problem here. The "thick wireframe" is the standard selection outline. Usually you'd see just the silhouette, but when something goes wrong... One of our shaders fails to compile when switching to material display mode. Also matcaps in solid display mode. Intel driver's GLSL compiler on Windows 10 complains: GPUShader: compile error: ERROR: 3:8: 'varying' : cannot be a struct ERROR: 3:27: 'varposition' : undeclared identifier ERROR: 3:27: 'assign' : cannot convert from '3-component vector of float' to 'float' ERROR: 3:28: 'varnormal' : undeclared identifier ERROR: 3:28: 'assign' : cannot convert from '3-component vector of float' to 'float' ERROR: 3:42: 'outpt' : undeclared identifier ERROR: 3:42: 'v' : field selection requires structure, vector, or matrix on left hand side ERROR: 3:42: 'position' : field selection requires structure, vector, or matrix on left hand side ERROR: 3:42: 'assign' : cannot convert from '4-component vector of float' to 'float' ERROR: 3:43: 'v' : field selection requires structure, vector, or matrix on left hand side ERROR: 3:43: 'normal' : field selection requires structure, vector, or matrix on left hand side According to grep, "varposition" is found in gpu_shader_[geometry|vertex|vertex_world].glsl

Mike Erwin commented

2015-11-14 22:37:09 +01:00

Reproduced problem on AMD A8-4555M running Windows 10. Shader compiler gave exact same errors as DBrown above.

Errors reported in my previous comment are from an Intel Haswell i5-4670.

Reproduced problem on AMD A8-4555M running Windows 10. Shader compiler gave exact same errors as DBrown above. Errors reported in my previous comment are from an Intel Haswell i5-4670.

Mike Erwin self-assigned this 2015-11-15 03:50:37 +01:00

Mike Erwin commented

2015-11-15 03:50:37 +01:00

Making some progress! Matcap and Cycles material preview are now working. No BI material preview yet. Haswell not working but for a different reason now. So it's not ready to commit yet. I will continue with this tomorrow.

Bumped GLSL from 1.3 to 1.5 (ARB_compatibility enabled) to get it working, plus changes to codegen and the OSD display shader.

{F254639}

Making some progress! Matcap and Cycles material preview are now working. No BI material preview yet. Haswell not working but for a different reason now. So it's not ready to commit yet. I will continue with this tomorrow. Bumped GLSL from 1.3 to 1.5 (ARB_compatibility enabled) to get it working, plus changes to codegen and the OSD display shader. {[F254639](https://archive.blender.org/developer/F254639/opensubdiv-material.png)}![opensubdiv-matcap.png](https://archive.blender.org/developer/F254643/opensubdiv-matcap.png)

Mike Erwin commented

2015-11-15 19:56:19 +01:00

Ok, got Intel Haswell working too. Changed

- version 150
- extension GL_ARB_compatibility : enable

which works fine on AMD to

#version 150 compatibility

which should work everywhere. GLSL spec says this is the proper way.

Basic uniform materials (for BI or Cycles) are working now in Texture and Material display modes. Anything with UV maps needs additional work, as @Sergey mentioned in the OpenSubdiv release notes.

Ok, got Intel Haswell working too. Changed ``` - version 150 - extension GL_ARB_compatibility : enable ``` which works fine on AMD to ``` #version 150 compatibility ``` which should work everywhere. GLSL spec says this is the proper way. Basic uniform materials (for BI or Cycles) are working now in Texture and Material display modes. Anything with UV maps needs additional work, as @Sergey mentioned in the OpenSubdiv release notes.

Sergey Sharybin commented

2015-11-18 12:11:50 +01:00

@MikeErwin, will you commit the fix?

Mike Erwin commented

2015-11-18 17:43:35 +01:00

@Sergey The current fix involves upgrading all codegen to GLSL 1.5. Which is great for Blender 2.8 but committing as is would break Mac builds and get in the way of 2.77. I'll try to make a minimal fix that is compatible with GLSL 1.2 and uses GLSL 1.5 (compatibility profile) just for the OSD parts.

I'll focus on this and should close before the weekend.

(except UV maps, that will take longer)

@Sergey The current fix involves upgrading all codegen to GLSL 1.5. Which is great for Blender 2.8 but committing as is would break Mac builds and get in the way of 2.77. I'll try to make a minimal fix that is compatible with GLSL 1.2 and uses GLSL 1.5 (compatibility profile) just for the OSD parts. I'll focus on this and should close before the weekend. (except UV maps, that will take longer)

Sergey Sharybin commented

2015-11-19 13:30:07 +01:00

@MikeErwin, sounds cool. And yes, don't touch UVs yet, it's more complicated topic and we can collaborate on this in IRC perhaps.

blender-admin commented

2015-11-23 09:35:16 +01:00

This issue was referenced by f997449f84

This issue was referenced by f997449f841f8229680f9e0f0beed4809b9111ab

Mike Erwin commented

2015-11-23 09:36:58 +01:00

Changed status from 'Open' to: 'Resolved'

Mike Erwin closed this issue

2015-11-23 09:36:58 +01:00

Mike Erwin commented

2015-11-23 09:42:36 +01:00

Changed status from 'Resolved' to: 'Open'

Mike Erwin reopened this issue

2015-11-23 09:42:36 +01:00

Mike Erwin commented

2015-11-23 09:42:36 +01:00

Partially resolved by recent commit.

Matcaps now work with OpenSubdiv, as do basic materials. Anything with UV coordinates is still broken.

+1 to working with @Sergey on UV part

Partially resolved by recent commit. Matcaps now work with OpenSubdiv, as do basic materials. Anything with UV coordinates is still broken. +1 to working with @Sergey on UV part

Brecht Van Lommel commented

2019-03-19 10:52:56 +01:00

Added subscriber: @brecht

Brecht Van Lommel commented

2019-03-19 10:52:56 +01:00

Changed status from 'Open' to: 'Archived'

Brecht Van Lommel closed this issue

2019-03-19 10:52:56 +01:00

Brecht Van Lommel commented

2019-03-19 10:52:56 +01:00

Archiving old report, we no longer work on fixing these types of issues in 2.79 but focus on the new viewport in 2.8.

Sign in to join this conversation.

No Label

Download

What's New

Blender Studio

Manual

Developers Blog

Documentation

Benchmark

Blender Conference

Development Fund

One-time Donations

Blender Render Texture and Material Views Show Large Wireframes when viewing OpenSubDiv Objects #46706