Weird Point Light Shadow Glitch

[marauder2k9]

http://i.imgur.com/gQSZJH1.png

Anyone got any ideas as to why this is happening?

I should say i increased the pre-taps in advanced lighting to 20 with closer poisson disks and it gives nicer edges to the shadows but i think this was happening before i did that too

[marauder2k9]

http://i.imgur.com/S4dVfUM.png

Shadows look really nice elsewhere just when ur standing in the middle of where the point light is placed

[marauder2k9]

http://i.imgur.com/zjSx2sh.png

rotating the light by 90 on its x axis fixed the problem, also fixed the problem with shadows on a few other objects too

[Jason Campbell]

That is a nice looking shadow!

[marauder2k9]

thank you :)

[JeffR]

Point Light shadows - when set with the Dual Parabloid shadow mode - have a known limitation/artifact in which the seam equator of the two hemispheres meet imperfectly, and a gap exists. This is due to the nature of trying to take 2 flat renders and use a warping algorithm to have it cover 180 degrees, naturally it's imperfect and prone to limitations/precision issues at the edges. A tradeoff from managing point light shadows with only 2 renders - it's a LOT faster, but obviously the limitations/artifact outlier cases can be unpleasant.

Nominally, you'd either rotate the light so the seam isn't noticed, or use the cubemap shadows, which while slower, don't have the seam issue.

Having said that, it could also be other odd precision issues or a different bug, so it's still worth testing on to try and confirm. So for starters, when you rotate the light, does that seam follow the rotation? Or does it disappear entirely?

[marauder2k9]

hey jeff,

yea the seam follows the lights rotation, rotating it 90 degrees works for my scene it hides the seam in a place that players cant really get to.on the cubemap shadows this project is only going to be released to 4 people as a kind of educational / fun activity, i couldnt get the cubemap shadows to render character model shadows, does a good job of architecture shadows hang on ill up screenshots

[marauder2k9]

this is the scene as is with dualparabloidsinglepass

http://i.imgur.com/Jmq86p2.png

and this is the scene with cubemap

http://i.imgur.com/W9TdihS.png

[JeffR]

Very odd! Good to document though. Especially weird because the walls clearly are shadowing alright. Wonder if it's a depth issue causing the player to be too close in the cubemap mode...

[marauder2k9]

I'll try a few things when I get home see if resizing the radius of the light fixes it, this was just a quick change to the light I had originally set up I had to rotate it back to original to get the shadows to display correctly

[marauder2k9]

I should also say that I'm gonna be changing this hallway to being a bit more modular but I've noticed that when two meshes join together at their edges it also causes a bit of a shadow glitch where shadows from one mesh are overlayed onto the next I'll get a screenshot when I get home, I think it was only happening when two lights where in the scene but a question I have is this, is there a way to get the suns light to shine through a surface with no backface, the ceiling in this example is only one polygon with some chamfering in the edges there's no backface to it it would be handier than using point lights and would probably produce better shadows too along with the ambient setting. I was using a sun in this scene just for the ambient setting with it's brightness set to zero and shadows set to off but it just seems too much of a waste to do that.

I also tried to use basic lighting instead, but none of my meshes are producing a shadow. I have created datablocks for a static shape to access its animations from a trigger. the old play thread execution. for opening doorways. but they don't produce a shadow in basic lighting neither does any tsstatic shape as far as i can remember

[RichardM]

If the player material is set to cast dynamic shadows, then they will not render as cubemap shadows, for the simple reason that dynamic cubemap shadows are not implemented in the shadow code.

[Azaezel]

https://gist.github.com/Azaezel/0a0bda20696d045e40ad478924b2c96a#file-gistfile1-txt-L113 + https://gist.github.com/Azaezel/0a0bda20696d045e40ad478924b2c96a#file-gistfile1-txt-L205-L208 should (but isn't) be all that was needed to get that squared away. Will have to do a bit more digging on that as time allows unless someone else has already been looking into that end of things for what we missed there.

[marauder2k9]

has it been tried with that if statement removed?