[x] Recently, we've noticed black pixel Z-fighting with r_depth_pass 1. This is easiest to reproduce by spawning into any map with an axially aligned spawn point, and simply walking around without changing your view angles. @Paril and I have both seen it on Edge, Lava Tomb, etc.
[x] Materials stages can sometimes "leak" through solid geometry. This is visible in several spots on Frag Pipe, where you'll see a hairline crack of the top of a lava surface through an adjacent wall.
Both 1 and 2 are likely an issue with our use of glPolygonOffset. I have never understood the intended usage of the offset and factor parameters to that function. Someone who does could probably fix both issues in short order.
[x] When standing on the cusp of an occlusion query, for some reason the query will snap in and out of pass / fail. This easiest spot to see this is to spawn into Lava Tomb here:
Just the small difference in view origin from cg_bob 1 is enough to teeter on the occlusion query. You can see in one shot, the func_plats, flares and fireballs are all missing and then present the next frame.
I've tried expanding the view origin bounding box when determining if we are inside an OQ. I don't know if the OQ is being culled the moment we're just outside of it, even tho it is still in front of us? I don't know what is causing this, but the flickering it produces happens often enough that we need to address it.
[x] Recently, we've noticed black pixel Z-fighting with
r_depth_pass 1. This is easiest to reproduce by spawning into any map with an axially aligned spawn point, and simply walking around without changing your view angles. @Paril and I have both seen it on Edge, Lava Tomb, etc.[x] Materials stages can sometimes "leak" through solid geometry. This is visible in several spots on Frag Pipe, where you'll see a hairline crack of the top of a lava surface through an adjacent wall.
Both 1 and 2 are likely an issue with our use of glPolygonOffset. I have never understood the intended usage of the offset and factor parameters to that function. Someone who does could probably fix both issues in short order.
Just the small difference in view origin from
cg_bob 1is enough to teeter on the occlusion query. You can see in one shot, thefunc_plats, flares and fireballs are all missing and then present the next frame.I've tried expanding the view origin bounding box when determining if we are inside an OQ. I don't know if the OQ is being culled the moment we're just outside of it, even tho it is still in front of us? I don't know what is causing this, but the flickering it produces happens often enough that we need to address it.