Speed up the coverage job (about 25 minutes since coverage became real)
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
- 1-2 天
- 新手友好度
- 55/100
- Issue 类型
- 缺陷
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
- 技术栈
- github-actions, python
- 领域
- ci-cd, performance, testing
调研方向
Start with the coverage job in ci.yml and inspect its step timings in the run for main, especially the analytic IK tests. Try COVERAGE_CORE=sysmon in the coverage step and compare the reported total with 85.98%. Done means the job runs faster and the coverage change is measured and explained; the issue also lists larger follow-up options if that first trial is insufficient.
由索引模型根据 Issue 内容生成。
描述
Since #897 made coverage real, the coverage job in ci.yml takes about 25 minutes (1,503 s on the run for main), where the other jobs take 10 to 15. It runs on every pull request and every push to main, since the job has no if:. It is not a required check, so it never blocks a merge, but it is the last job to finish on every run, and the one that tells the Codecov number.
Why it is slow
It runs the whole suite with coverage tracing, including the tests marked slow (the analytic IK tests that compile generated C and C++ code, about 110 s untraced), and it runs them under the default tracing core.
Ideas, cheapest first
-
COVERAGE_CORE=sysmonin the coverage step. Python 3.12 has a lighter tracing core. Try it on the job, and compare the totals with the current 85.98%, since a different core can change which lines count. - Do not trace the slow tests: run them in the matrix job on Linux with no coverage, and take coverage from the fast tests only. Check what the analytic IK tests contribute (they exercise about 4,000 lines).
- Split the job with
pytest-xdistand combine the data (coverage combine), as the tests are independent. - Run coverage on pushes to
mainand weekly, not on every PR. The PR would lose its Codecov patch status, which is informational now anyway. - Cache what is rebuilt on every run: the package install,
robot_descriptionsassets (already cached).
Measure first
The job log has the timing for each step. Note the compile-based tests in particular: on the Windows runner they were slow enough to hit the 50 s limit, so a Linux job that traces them may be paying the same cost.
Related: #894 (code-quality follow-ups), petercorke/rvc-ecosystem#26 (the CI audit that found the coverage problem).
- 主要语言
- C++
- 星标
- 3.6k
- 派生
- 645
- 平均合并
- 1 天 8 小时
- 30 天内合并 PR
- 145
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。
petercorke/robotics-toolbox-python 的其他 Issue
-
plt.pause(0) waits for ever: PoseGraph.plot_scan(pause=0) and Vehicle animation may hang可能已有人在做 @huxiaoyi-ovo 今天认领。 未关闭bug
难度 2/5 1-3 小时 新手友好度 72/100
petercorke/robotics-toolbox-python#876 · 1 条评论 · 已指派 1 人 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 68/100
petercorke/robotics-toolbox-python#709 ·
维护者通常 1 天内回复
-
enhancement
难度 5/5 一周以上 新手友好度 25/100
petercorke/robotics-toolbox-python#903 · 1 条评论 ·
维护者通常 1 天内回复
-
Backend registry: single name → constructor dispatch可能已有人在做 @petercorke 今天认领。 未关闭enhancement
petercorke/robotics-toolbox-python#902 · 已指派 1 人 ·
维护者通常 1 天内回复
-
LatticePlanner: integer headings, and motion primitives with finer turn steps (design first)可能已有人在做 @kVinsom 于 3 天前认领。 未关闭enhancement help wanted
难度 5/5 一周以上 新手友好度 25/100
petercorke/robotics-toolbox-python#895 · 2 条评论 ·
维护者通常 1 天内回复
查看 petercorke/robotics-toolbox-python 的全部 Issue
相似的 Issue
-
[Bug] [Go] Ack returns nil when the server answers with a non-OK status可能已有人在做 @leizhiyuan 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 84/100
apache/rocketmq-clients#1465 ·
维护者通常 2 天内回复
-
难度 2/5 1-3 小时 新手友好度 72/100
KhronosGroup/Vulkan-Tutorial#534 ·
-
难度 2/5 1-3 小时 新手友好度 65/100
维护者通常 2 天内回复
-
bug IO
难度 2/5 1-3 小时 新手友好度 65/100
MrNeRF/LichtFeld-Studio#3824 ·
维护者通常 1 天内回复
-
pnnx: manual tanh GELU fusion loses the approximation mode可能已有人在做 @ktz03 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 78/100
Tencent/ncnn#7175 · 1 个 reaction ·
维护者通常 1 天内回复