Python adapter test: interpreter probed twice can disagree on Windows (python3 vs python), failing CI at random

未關閉 適合新手
#304 0 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視

維護者通常 1 天內回覆

還沒有人認領這個 Issue。

評估

難度
2/5
預估耗時
1-3 小時
新手友好度
78/100
Issue 類型
缺陷
描述清晰度
描述清楚
活躍度
活躍
技術堆疊
github-actions, python, typescript
領域
ci-cd, testing-qa

研究方向

從 tests/repo/python-adapter.test.ts 開始,特別查看第 111 行附近的直譯器斷言,接著閱讀 src/repo/adapters/python.ts 以及第 70 行附近測試中的直譯器設定。執行 Python 配接器測試,並檢查它是否與分析期間記錄的直譯器進行比對,或者是否能容忍因工作目錄不同而使用不同的 Python 3 啟動器。完成標準是 Windows 上的失敗不再依賴兩次分開的直譯器探測。

由索引模型根據 Issue 內容生成。

描述

tests/repo/python-adapter.test.ts ("selects the pytest files a change reaches and runs them with the interpreter it found") failed once on check (windows-latest, 24) for the docs-only PR #291 (run 44200296961, job 134452346206). The other three check jobs in that run passed:

AssertionError [ERR_ASSERTION]: Expected values to be strictly deep-equal:
+     executable: 'python3'
-     executable: 'python'
    at TestContext.<anonymous> (D:\a\core\core\tests\repo\python-adapter.test.ts:111:14)

Why it can disagree. The test and the adapter each probe for an interpreter, in different directories and at different times:

  • The test computes the expected interpreter once at module load: const interpreter = pythonInterpreter(process.cwd()); (L70).
  • The adapter calls pythonInterpreter() again, with the fixture's temp directory as cwd, when it builds the graph.
  • pythonInterpreter() (src/repo/adapters/python.ts L74) tries python3, then python, then py -3, each with a 10 s spawnSync timeout, and takes the first that prints 3.

On Windows runners, python3 is often the App Execution Alias stub or a slow first start. One probe can fail or time out on python3 and fall through to python, while the next probe succeeds with python3. The command the adapter writes is valid either way, but the assertion pins the test's own probe result.

This is not the known Rust temp-dir EBUSY/EPERM flake (#235); the failing test and the error are different. Recent Windows failures on other PRs were all the Rust one, so this looks rare.

Expected: the test asserts the interpreter the analysis recorded (or any working Python 3 launcher), not a second independent probe. Alternatively, pythonInterpreter() gives the same answer for the whole process, so one run cannot get two different interpreters.

主要語言
TypeScript
星號
0
分支
4
平均合併
18 小時 41 分鐘
30 天內合併 PR
48

環境準備

在 Codespaces 中開啟

在瀏覽器裡用你自己的 GitHub 帳號啟動這個專案的開發容器。

  • 沒有 Dockerfile 或 Docker Compose 檔案
  • 沒有 Pull Request 範本
  • 閱讀貢獻指南

從這裡開始

  1. 先讀完整個 Issue,再讀專案的貢獻指南。
  2. 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
  3. Fork 儲存庫,在一個分支上完成修改。
  4. 送出 Pull Request,並在描述裡引用這個 Issue 編號。

DiffCI/core 的其他 Issue

查看 DiffCI/core 的全部 Issue

相似的 Issue

更多 TypeScript Issue

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。