Selection bench: main-thread stall p95 13 ms over its 8 ms budget on macOS (never held there)
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 新手友好度
- 35/100
- Issue 类型
- 缺陷
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
- 领域
- desktop, performance
调研方向
Start by reading scripts/selection-bench.ts and run bun scripts/selection-bench.ts on the release Mac to reproduce the stall and inspect its measurements. The issue points to time after the synchronous flush, including WebKit style, layout, and paint, but does not identify the code responsible. Done means reducing p95 stall below 8 ms or setting a budget from a measured release Mac baseline; desktop UI work is owned outside the crew per #348.
由索引模型根据 Issue 内容生成。
描述
Selection bench: main-thread stall p95 is over its 8 ms budget on macOS (and was at the commit that set it)
Found in the macOS release checks for 1.5.0-alpha.26 (bun scripts/selection-bench.ts, agent instance, 13 fixture repos, 40 rounds, idle machine: load ~2.5 on 12 cores).
| stall p95 | flush p95 | nodes p95 | |
|---|---|---|---|
budget (selectionBench.ts) |
8 ms | 8 ms | 120 |
bba68598 (2026-10-13, the commit that set the budget) |
11 ms | — | — |
main c48f7dc9 (3 runs) |
13 / 13 / 14 ms | 5 ms | 22 |
So this is not a new regression: the stall budget never held on this Mac (M4 Pro, macOS 27.2). It drifted ~2-3 ms in a month while flush (4→5 ms) and nodes (84→22) improved. Flicker sweep on the same build: 19 surfaces, 0 blips, 0 blank, 0 ghost frames.
To decide: find what fills the ~13 ms gap after the flush (the stall is measured by a MessageChannel ping-pong, so it includes WebKit style/layout/paint after the synchronous flush, paintMs p95 is 38 ms) and bring it under 8 ms, or set the budget from a measured baseline on the release Mac. Desktop UI area (Linux agent owns apps/desktop outside crew per #348), filed for triage; not a .26 blocker.
- 主要语言
- Rust
- 星标
- 3
- 派生
- 1
- 平均合并
- 9 小时 51 分钟
- 30 天内合并 PR
- 310
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue,再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库,在一个分支上完成修改。
- 提交 Pull Request,并在描述里引用这个 Issue 编号。
codeatlasdev/atlas 的其他 Issue
-
type: bug
难度 2/5 1-3 小时 新手友好度 75/100
codeatlasdev/atlas#597 ·
维护者通常 1 天内回复
-
area: ci os: linux type: bug
难度 2/5 1-3 小时 新手友好度 78/100
codeatlasdev/atlas#496 ·
维护者通常 1 天内回复
-
area: ci type: bug
难度 3/5 1-2 天 新手友好度 68/100
codeatlasdev/atlas#596 ·
维护者通常 1 天内回复
-
os: linux type: bug
难度 3/5 1-2 天 新手友好度 68/100
codeatlasdev/atlas#591 ·
维护者通常 1 天内回复
-
area: ci os: windows type: bug
难度 3/5 1-2 天 新手友好度 56/100
codeatlasdev/atlas#550 · 1 条评论 ·
维护者通常 1 天内回复
查看 codeatlasdev/atlas 的全部 Issue
相似的 Issue
-
marketing
难度 2/5 1-3 小时 新手友好度 76/100
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 74/100
kinqsradio/orbyn-engineering#729 ·
-
Plans page scenario races Shift+Tab against the plan list being drawn again in Firefox and WebKit未关闭bug
难度 2/5 1-3 小时 新手友好度 83/100
nledford/Ladderback#1091 ·
维护者通常 1 天内回复
-
bug fix packaging ui user-sprint
难度 2/5 1-3 小时 新手友好度 78/100
Vondor/KassenSturz#1197 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 82/100
维护者通常 1 天内回复