cesium1.36 Model transparency

[duoduohappy]

cesium1.36

cesium1.35
Find a problem, cesium1.36 The model becomes transparent,cesium1.35 Display normal

[lilleyse]

Would you be able to send over the model? That will help for debugging this.

[pmconne]

I'm also observing this issue. When globe is on, depth test appears to be disabled for translucent geometry only. If you have our OfficeBuilding data set, that one exhibits it - or I can send a simpler test case.

[pmconne]

Translucent.zip

Globe off (translucent red cylinder behind opaque green cylinder):

Globe on:

[lilleyse]

Thanks @pmconne for uploading the model, I'll start looking into this and #5734.

[lilleyse]

@pmconne It seems like this is caused by the recent architectural changes that allowed classification primitives to work on 3D Tiles - see #5676 for more context.

Now the globe and opaque 3D Tiles are rendered together into the same depth buffer, and their depth is cleared after classification primitives are rendered. This causes the red cylinder, which is in the translucent pass, to not depth test against anything.

For the time being our recommendation is to set scene.globe.depthTestAgainstTerrain to true, but we hope to have a better solution before the next release.

[duoduohappy]

Thanks @lilleyse , scene.globe.depthTestAgainstTerrain=true Display normal

[duoduohappy]

@lilleyse vector layer Will the next release increase?

[lilleyse]

@lilleyse vector layer Will the next release increase?

It's unlikely that vector tiles will be ready by the next release. Or do you mean something else?

[duoduohappy]

Thanks @lilleyse We refer to mapbox vector label and vector line , I wonder if Cesium will increase next release？

[pjcozzi]

@lilleyse please close if you confirmed this is fixed in #5770.

[pjcozzi]

Reopening to investigate @pmconne's use case when the globe is off.

[pjcozzi]

Pushing post 1.41 release. @bagnell please still look at this soon, thanks!

[pjcozzi]

@bagnell could you please look at this soon when you need a break from tests in #5615?

[bagnell]

I can't reproduce this. I see the correct rendering regardless of whether the globe is on or off.

Translucent in front:

Translucent in back:

[hpinkos]

@bagnell this area looks questionable

[hpinkos]

Or wait, I guess they're intersecting so that would be correct

[hpinkos]

I'm going to close this. I got the same results @bagnell did