Bugs Related to PBS


#1

Putting this thread up so that people can put all the bugs related to PBS here instead of posting multiple threads on issues they have:

JPG Rougness vs (super)PNG Roughness as test textures.

The above could be related to compression settings. The PNG was exported using SuperPNG, instead of defaults used by Photoshop. This could lead to frustration, and should probably be checked, as superPNG can be used to compress or clean up png textures instead of using jpgs.

CC: @chris and @sam


#2

Can you tell me where in the domain you were when you tested that? It may be related to a known issue with floating point drift issues as you move away from the origin.


#3

I was in my domain Center (points towards the right coordinates). The object in question has already been rezed.


#4

has observed that in 3rd person mode if u scroll the mouse scroll wheel the shadows move even if you don’t.

In rift mode I’m getting jittery jumpy performance and no shadows @63fps in 4319


#5

As I mentioned somewhere before, this sort of breaks any use of polyvoxels. At best they are just colored blobs that change color as the angle of view changes. At worse they disappear completely.


#6

This seems to also occur at 0,0,0 of the domain.


#7

More related bugs:

  • If object spins, the reflected environment spins as well
  • Mirrored Geometry with same UV maps normals (this was reported earlier, but now making sure it reaches) have buggy normal direcitons
  • Extra notes: Seems like if an object doesnt have defined normal maps, it works as intended (exception being jitter). (see Center) for live examples.

CC: @sam sorry :stuck_out_tongue: Great work so far but I keep bumping into these XD


Support for mirrored UV's / Normals?
#8

@Samcake I really do like the new PBR stuff, when it works it looks much better than old way. It will be a chore to fix lots of stuff that is here already, but most of what is here could use an update anyway.

As @Menithal has said, there are still a few problems, some which have work arounds. What this post is about is I spent a long time trying to reproduce some of the stuff I knew would be possible here.

After 23 tries trying to replicate what I had seen others doing here on the 24th try, I finally tried using a 32bit .png instead of a .jpg for some of the textures and YES… it finally worked.

Not sure what the difference is really, This was not a texture that I thought would use transparency channel but anyway, I only post this as a hint to anyone having similar troubles.

So not really a bug or anything. Yaay for PBR and all is well in Alpha land.


#9

I can replicate the bugs on my work Macbook.

What I have to say is impressive though, this is Running at 60fps on my new Macbook Pros just fine and equally pretty as on my desktop. (Mid 2015, 15 Inch retina, 2.2 and Iris Pro 1536 MB)


#10

is that loaded right? has a doubt i did it correctly


#11

It looks like it is missing the roughness. Check using the debugFramebuffer script: its comes with the new version.

Otherwise try making roughness into jpg or 32bit png without transparency.


#12

Not quite right.
Color 1.0
Normal Normal Map Normal 1.0
AO Har 1.0
Specualr Inte 1.0


#13

Did you try the debugFramebuffer to double check what textures bound to what?


#14

That sounds like @judas ring buffer sounds like the work of the devil.

Not sure what your referring to do you have an example?


#15

it is a js script that should be under scripts/utilities/tools/render/debugFramebuffer.js It is really handy for checking what texture is behaving incorrectly. if it is a texture at all or if the issue is caused by something else entirely.


#16

A New bug :smiley:

If alpha map is defined, Glossiness texture does not get applied at all:

Meaning we cant have both alpha and glossiness/rougness definitions.

Nope, Apparently it is caused by the darn compresssion/png loading issues. Got it to work after switching roughness to a jpg.


#17

Transparency Bugs:

If the object is of the same mesh, transparent objects blend fine
http://www.norteclabs.com/HF/experiment/transparencytest4.fbx

If objects is of different mesh, but of the same object, transparent objects have conflicts. This effects also when they are separate objects in Hifi.
http://www.norteclabs.com/HF/experiment/transparencytest5.fbx

Setup available at Center.

CC: @chris


#18

The difference between the two photos is me moving slightly left and right relative to the spheres. Note also that the pair of spheres to the left do not exhibit the alpha conflict when I move about. It looks like there is some order of operations difference going on.

[correction] the difference being that one .fbx model was two objects, each made out of a sphere while the other .fbx model was one object made out of two spheres.


#19

The way I understood it was that they both were single .fbx files, the difference being that one .fbx model was two objects, each made out of a sphere while the other .fbx model was one object made out of two spheres.


#20

Yes, I misunderstood and corrected my comments.