LichtFeld .rad chunks at 2048 splats/chunk → ~67k chunks inlined into a ~8 MB RAD header. Spark's own build-lod uses 65536 (~2k chunks → ~250 KB header).
| # |
Blocker |
Detail |
| 1 |
Header too large |
Spark's SplatPager reads ≤1 MB for the header (fileBytes.slice(0, 1048576); URL path tries 64 KB → 256 KB → 1 MB, then throws Failed to decode RAD header). LichtFeld's header is ~8 MB. |
| 2 |
LOD-tree / chunk layout |
After locally raising the 1 MB cap, Spark's WASM panics: lod_tree.rs:531 index out of bounds: the len is 2048 but the index is 2811 — the chunk/LOD-tree binary layout also diverges. |
Verified on Spark 2.1.0 (npm latest; no next/main published). The 1 MB cap is in both 2.1.0 and current main, so main won't load it either; blocker #2 was seen on the 2.1.0 runtime (on main it fails at #1 first).
LichtFeld
.radchunks at 2048 splats/chunk → ~67k chunks inlined into a ~8 MB RAD header. Spark's ownbuild-loduses 65536 (~2k chunks → ~250 KB header).SplatPagerreads ≤1 MB for the header (fileBytes.slice(0, 1048576); URL path tries 64 KB → 256 KB → 1 MB, then throwsFailed to decode RAD header). LichtFeld's header is ~8 MB.lod_tree.rs:531 index out of bounds: the len is 2048 but the index is 2811— the chunk/LOD-tree binary layout also diverges.Verified on Spark
2.1.0(npmlatest; nonext/mainpublished). The 1 MB cap is in both 2.1.0 and currentmain, somainwon't load it either; blocker #2 was seen on the 2.1.0 runtime (onmainit fails at #1 first).