Linux - Removing point entity with HeightReference.NONE causes a phantom entity to appear

[floonone]

Removing a point entity with the height reference property set to NONE causes a phantom entity to appear in the middle of the screen. This entity has no id and doesn't show up in the collection, so it can't be removed, unless a new point entity is added to the collection, which makes it go away.

Depending on the camera position, the entity disappears and reappears. Setting heightReference to CLAMP_TO_GROUND or RELATIVE_TO_GROUND works as expected.

Sandcastle example

Thanks in advance!

[hpinkos]

Hi @floonone, thanks for including the Sandcastle example! However, everything seemed to work as expected for me. Can you paste a screen shot of what you're seeing?

[floonone]

Here it is

I've managed to reproduce it in Firefox 62 and Chrome 69 on Ubuntu 16.04, but wasn't able to in Firefox 62 on Windows 10. Could it be a driver or a WebGL misconfiguration issue?

[hpinkos]

Hmm. Yeah, I'm running Windows and couldn't reproduce this. If your drivers are out of date, that's one thing that could cause an issue.

@mramato can you see if you can reproduce this on your linux machine?

[mramato]

I can't seem to reproduce this issue either, I tried both Firefox 62 and Chrome 69 on Kubuntu 18.04. I agree this could be a graphics driver, especially since the point gets all floaty and detached from the scene. Is there any output in the console? What video card are you using?

Can you visit http://webglreport.com/ on Chrome 69 on the offending machine and paste the results of the WebGL 1 report (just select all/copy and GitHub will make a pretty table when you paste it).

[floonone]

Is there any output in the console? What video card are you using?

No console output. I'm using a Nvidia GeForce GT 730, using Nvidia binary driver version 384.130.

Can you visit http://webglreport.com/ on Chrome 69 on the offending machine and paste the results of the WebGL 1 report (just select all/copy and GitHub will make a pretty table when you paste it).

Here it is:

Info
Platform:	Linux x86_64
Browser User Agent:	Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/69.0.3497.81 Safari/537.36
Context Name:	webgl
GL Version:	WebGL 1.0 (OpenGL ES 2.0 Chromium)
Shading Language Version:	WebGL GLSL ES 1.0 (OpenGL ES GLSL ES 1.0 Chromium)
Vendor:	WebKit
Renderer:	WebKit WebGL
Unmasked Vendor:	NVIDIA Corporation
Unmasked Renderer:	GeForce GT 730/PCIe/SSE2
Antialiasing:	Available
ANGLE:	No
Major Performance Caveat:	No

Supported Extensions
ANGLE_instanced_arrays
EXT_blend_minmax
EXT_color_buffer_half_float
EXT_frag_depth
EXT_shader_texture_lod
EXT_texture_filter_anisotropic
WEBKIT_EXT_texture_filter_anisotropic
EXT_sRGB
OES_element_index_uint
OES_standard_derivatives
OES_texture_float
OES_texture_float_linear
OES_texture_half_float
OES_texture_half_float_linear
OES_vertex_array_object
WEBGL_color_buffer_float
WEBGL_compressed_texture_s3tc
WEBKIT_WEBGL_compressed_texture_s3tc
WEBGL_compressed_texture_s3tc_srgb
WEBGL_debug_renderer_info
WEBGL_debug_shaders
WEBGL_depth_texture
WEBKIT_WEBGL_depth_texture
WEBGL_draw_buffers
WEBGL_lose_context
WEBKIT_WEBGL_lose_context

Vertex Shader
Max Vertex Attributes	16
Max Vertex Uniform Vectors	1024
Max Vertex Texture Image Units	32
Max Varying Vectors	31
Best float precision	-2127, 2127

Rasterizer
Aliased Line Width Range	[1, 10]
Aliased Point Size Range	[1, 189.875]

Fragment Shader
Max Fragment Uniform Vectors	1024
Max Texture Image Units	32
float/int precision	highp/highp
Best float precision	-2127, 2127

Framebuffer
Max Color Buffers	8
RGBA Bits	[8, 8, 8, 8]
Depth / Stencil Bits	[24, 8]
Max Render Buffer Size	16384
Max Viewport Dimensions	[16384, 16384]

Transform Feedback
Coming in WebGL 2

Textures
Max Texture Size	16384
Max Cube Map Texture Size	16384
Max Combined Texture Image Units	192
Max Anisotropy	16

Uniform Buffers
Coming in WebGL 2

[hpinkos]

Also reported by @cpt-chewieeee in #7045. See #7045 for an example using clustering.

@cpt-chewieeee was also seeing this issue on a linux machine, see #7045 (comment) for more details

[hpinkos]

@bagnell do you have a linux machine you can try to reproduce this on?

@tfili do you still have your linux machine?

[supercamel]

Hi,

I have also encountered this issue on my linux machine using Electron.
Version string from Electron is . .
{ http_parser: '2.8.0', node: '10.2.0', v8: '6.6.346.32', uv: '1.20.3', zlib: '1.2.11', ares: '1.14.0', modules: '64', nghttp2: '1.29.0', napi: '3', openssl: '1.1.0h', electron: '3.0.0-beta.3', chrome: '66.0.3359.181' }
Cesium version is 1.51

Cheers

[hpinkos]

Thanks @supercamel! I'm not sure how soon we're going to be able to look into this, but we'll give you and update when we can

[UniquePanda]

As I stumbled over the same problem, just some system info from me, too:
Using Ubuntu 18.04 with Chrome 70.0 and latest available graphics drivers from Nvidia (415.18) with a GeForce GTX 660 Ti.

I can provide some vague additional information:
We use point entities with some functionality attached, so we check with pick/drillPick methods if a point entity is clicked. The "phantom points" seem to still trigger this, as the whole website stops working when I click one of them. Unfortunately I can't narrow this down and provide an example right now because I don't know what really is causing this.
I also managed to hide the "phantom points" alongside other primitives by setting the corresponding boolean to false (in viewer.scene.primitives).
So, long story short: It seems to not just be a graphical glitch. Cesium seems to be able to detect that something is there and the "phantom points" are still somehow connected to the scenes primitives.

Additionally the "phantom points" sometimes disappear, depending on which part of the globe I am looking and how far/near I zoomed in. Here's an example gif I already posted in the issue (#7381) I created before finding this one. In this example the "phantom point" is only appearing after zooming in and out:

[UniquePanda]

As I started to dig around a bit, I'll just leave my two cents here. Maybe it can at least give some tips if someone else wants to work on it (as I don't have the time right now unfortunately) ;)

1. Point entites without a height reference are handled as PointPrimitives, while point entities with a height reference are handled as Billboards. This is very likely the reason for the different behaviour of points that is mentioned here.

2. I think the problem does not lie in the methods to remove the entities. More likely it really is some problem in the rendering part (or the drivers).
   One thing I've noticed is, that the phantom entity does not appear when removing point.show = false in EntityCluster.prototype.removePoint() from Source/DataSources/EntityCluster.js, instead the point will be shown where it was before on the globe as if it hasn't been removed (which is expected behaviour, i think).
   I also modified the PointPrimitiveCollection Vertex Shader (PointPrimitiveCollectionVS.glsl). The phantom entity is influenced by some shader values (e.g. hard coding the color to green will make the phantom entity green) while it's not influenced by some other values (e.g. hard coding show to 0 will make the real entity disappear but the phantom entity is still displayed after removing the real one).
   I am guessing that the phantom entity is influenced by the values that are outputted by the vertex shader.
   Edit: Even though it looks like there is no primitive connected to the phantom entity, calling viewer.scene.primitives.show = false makes the phantom entity disappear.

3. viewer.scene.pick() and viewer.scene.drillPick() return undefined/an empty array when called with the screen position of the phantom entity (as expected when there really is no primitive anymore). But viewer.scene.pickPosition() also returns undefined on this screen position (which is causing the problem I described in my comment above, i think). So the phantom entity at least is messing with the depth buffer (depth is calculated as 0).

[hpinkos]

Thank you for this detailed write up @UniquePanda! That should help our graphics developers narrow in on the problem.

[mramato]

I can absolutely reproduce this on Linux (Nvidia) and also via Cesium ion (the tileset position tool ends up with a large phantom point).

[afroman-AK]

I don't think removing the point in necessary to recreate the issue. In our application we were able to cause this by changing the 'show' variable.

I am still testing it but a possible work-around is to set the color to have an alpha of 0.0 and hide it via that method. Initially, that seems to be working.

[mramato]

This is definitely a video driver specific bug/issue on linux, which means it may not be our fault (or possible to workaround). Hopefully that's not the case but we need to dig in further to know for sure. The reason I know this is because I can reproduce it every time under the same hardware if I toggle back and forth between my intel and nvidia GPUs.

[lilleyse]

This was fixed in #8054 which went out with the 1.61 release.