2026 ED138 from the Rubin seat at a 0.00016° field: the disc blinks, the ring jumps, the path and the orbit line miss (float32 at 47 au)
メンテナーはふだん 1 日以内に返信
まだ誰も着手していません。
評価
- 難易度
- 5/5
- 見積もり時間
- 1週間以上
- 初心者へのやさしさ
- 35/100
- issue の種類
- バグ
- 明瞭さ
- 明確に書かれている
- 活発さ
- 活発
- 技術スタック
- javascript, three.js
調査の方向性
Read js/scene/rubin/Rubin.md, HDR.md and DESIGN.md for the seat, far point and LOD hand-off; reproduce the reported view near 47 au at a 0.00016° field. Compare the mesh, path and orbit line precision and depth behavior across the proposed approaches. Done means choosing and implementing a fix that keeps the disc, path and orbit line stable and correctly occluded at close zoom.
索引モデルが issue の本文から書いたものです。
説明
From the Rubin seat (#327, #328), zoomed with the field-of-view keys to about 0.00016° (0.6″ across the frame), on the maintainer's GPU (the preview of #320, 2026-10-28). Not reproduced headless on SwiftShader at pixel ratio 1 or 2: there the planet LOD is at level 0 (the mesh), the far point hidden, and the disc draws whole.
What it looks like (the maintainer's captures)
The clock paused at 256×, the night of 15 May 2026, a few steps of time apart:
- Gone: the disc not drawn at all; the path's ring well off the orbit line.
'https://celestiary.org/web/pr-preview/pr-320/#sun/[email protected],-70.7494,1.689km;from=sun/earth;t=9630.794665jd;cq=0.7206,0.557,0.0889,-0.4032;fov=0.0001615deg;s=LTF;time:pause,rate=-256;apps:open,view=rubin'
- A crescent: part of the disc, the dark bite with a curved edge.
'https://celestiary.org/web/pr-preview/pr-320/#sun/[email protected],-70.7494,1.689km;from=sun/earth;t=9630.806745jd;cq=0.7312,0.5532,0.1064,-0.3846;fov=0.0001615deg;s=LTF;time:pause,rate=256;apps:open,view=rubin'
- Whole: the disc drawn, the blue orbit line across its face, not occluded by it.
'https://celestiary.org/web/pr-preview/pr-320/#sun/[email protected],-70.7494,1.689km;from=sun/earth;t=9630.807563jd;cq=0.7319,0.5529,0.1076,-0.3834;fov=0.0001615deg;s=LTF;time:pause,rate=256;apps:open,view=rubin'
Also at that field: the green apparent path doesn't meet the blue orbit line; the path's ring (the body's place at the clock's time) jumps about between frames while everything else moves smoothly. Zoomed out to 1° the same view is fine.
Where it lives
A 500 km body at 47 au is one part in 1.4e7 of its distance, past float32's 2^23. Everything at 7e12 m from the camera is held to float32's spacing there, 5e5 m, which is 7e-8 rad, 0.015″: at a 0.00016° field that is 50 to 200 px.
- The mesh: its vertices land to within half the body's radius, differently per GPU and per frame as the camera moves (time running), so the disc jumps, loses parts, or isn't drawn (1, 2).
- The path and the ring: the path's points are a
Float32Arrayof positions at the body's distance (since #320 put the path at its true distance), quantised to 5e5 m, and the ring interpolates between two of them: the jump between frames is the quantisation stepping as the clock moves. The orbit line's vertices are float32 in the Sun's frame, quantised the same way but not on the same steps, so the two lines miss each other by up to a step. - Depth: the body's depth from the seat, 1 − 169 m / 7e12 m, is 1.0 in a 24-bit buffer, the same as empty sky, so nothing occludes against it: the orbit line draws across the disc (3), and the passes that read depth (the atmosphere's night sky, the galaxy's pin) can't tell the body from the sky either.
Pluto at 39 au and 1,200 km is just inside the limit; no other body reaches it.
That field is 20× past Rubin's own pixels
Rubin's pixels are 0.2″; the view at 0.6″ across the frame resolves nothing real, and the body's 0.03″ disc is a point to any telescope on Earth. Options:
- a field-of-view floor from the seat (or landed) at the instrument's resolution, a few arcseconds, where none of this shows (the maintainer's 1° frame is right);
- or keep the far point for a body whose distance over its radius passes 2^22, and draw no mesh there (the lines and the ring still quantise);
- or relative-to-eye placement of far things: the body's centre, the path's and the orbit line's points held in float64 on the CPU and uploaded as offsets from the camera each frame, which is the general fix and the costliest.
Design notes: js/scene/rubin/Rubin.md (the path, the seat), HDR.md (the far point), DESIGN.md (the far point and the LOD hand-off).
- 主要言語
- JavaScript
- スター
- 55
- フォーク
- 4
- 平均マージ
- 10時間 12分
- マージ済み PR（30日）
- 134
環境構築
このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
celestiary/web のほかの issue
-
難易度 1/5 1時間未満 初心者へのやさしさ 88/100
celestiary/web#345 ·
メンテナーはふだん 1 日以内に返信
-
Cesium guest (?cesium=iframe): white flicker near Mars's and Earth's surface remains after #335オープン
難易度 4/5 3〜5日 初心者へのやさしさ 35/100
celestiary/web#344 ·
メンテナーはふだん 1 日以内に返信
-
難易度 5/5 1週間以上 初心者へのやさしさ 25/100
celestiary/web#340 ·
メンテナーはふだん 1 日以内に返信
-
難易度 4/5 1〜2日 初心者へのやさしさ 48/100
celestiary/web#336 ·
メンテナーはふだん 1 日以内に返信
-
enhancement track:spacecraft
難易度 5/5 1週間以上 初心者へのやさしさ 28/100
celestiary/web#333 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信
似ている issue
-
FEATURE REQUEST
難易度 2/5 1〜3時間 初心者へのやさしさ 68/100
linuxmint/cinnamon-spices-extensions#1193 ·
メンテナーはふだん 1 日以内に返信
-
bug
難易度 2/5 1〜3時間 初心者へのやさしさ 76/100
Fei2411/DeepSeek-Bot#703 ·
-
area:web first-timers-only good first issue
難易度 1/5 1時間未満 初心者へのやさしさ 95/100
career-ops-hq/career-ops#6264 ·
メンテナーはふだん 1 日以内に返信
-
難易度 1/5 1時間未満 初心者へのやさしさ 74/100
catppuccin/firefox#21 ·
-
難易度 1/5 1時間未満 初心者へのやさしさ 83/100
cline/mcp-marketplace#3394 ·