random seed: an edit sequence through the daemon: authored mutant a2 runs to its backstop (timeout, owed killed), seed 964713679 on #99

未關閉
#167 1 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
5/5
預估耗時
一週以上
新手友好度
15/100
Issue 類型
缺陷
描述清晰度
基本清楚
活躍度
活躍
技術堆疊
javascript, vite
領域
testing-qa

研究方向

Start with the e2e oracle command and seed in the counterexample, then inspect the authored mutants law at the cli checkpoint and the history details in the issue. The report identifies a timeout in file b's authored:a2 copy, but the cause is still unknown and history replay depends on #157. Done means reproducing and explaining the timeout; the issue assigns this investigation to #99's author for a later follow-up round.

由索引模型根據 Issue 內容生成。

描述

Counterexample

A random seed of the e2e oracle on #99 at 83a0b3a8 (pool, then twice more, and on macOS: reproducible):

ORACLE_SEED=964713679 ORACLE_PATH=0 ORACLE_PROPERTY='oracle-gaps.report' ORACLE_UNIT='random seed: an edit sequence through the daemon' GAP_E2E_SHRINKS=0 vitest run --config oracle-gaps/vitest.config.mjs gap-report-oracle -t '^oracle-gaps\.report(?: | > )random seed: an edit sequence through the daemon$'
  • Law authored mutants, at checkpoint cli (the cold run): the authored mutant a2 is timeout, owed killed. Its result is {"outcome":"timeout","cause":"backstop"}.
  • File b's copy for authored:a2 runs to its backstop: 20.05 s, where the limit is 20 s plus 3 s × 0.018 s. Its file's unmutated run takes 0.018 s.
  • The history (cut at 2,000 characters in the log) has:
    • two functions, if x > 4 push and push, owned by a and b;
    • a test of file a with input 5, and a test of file b with input 1 on f1;
    • authored: {stmt: 0, kind: apply}, secondAuthored: true, ordinary: true, contexts: 4, budget: one;
    • cliPath: spaced, unbuiltDep: true, packageType: commonjs, configFile: vite.config.cts, concurrentStart: true.
  • The same seed passes on main 97a4e7a, but there it draws a different history, because #99 shifts the draws. So base attribution needs a replay by history (#157), which is not built yet.

What is known

  • The copy is not idle by state. If it were, it would be killed by no progress. It also has no call budget (an authored copy counts no calls).
  • The copy either loops synchronously, or waits before any test starts. Before a test starts the sampler does not judge idle; an example is a module fetch over the runner's RPC (#148).
  • It is not caused by #99's clocked-by-state hook: with that hook's completion record disabled, the unit fails the same way.

Owner

#99's author, in the next round on the deterministic-cuts follow-ups. Replay the full history on main once #157 lands; if main fails it too, this does not block #99 (M, "Main's determinism failures").

主要語言
JavaScript
星號
5
分支
1
平均合併
11 小時 18 分鐘
30 天內合併 PR
66

環境準備

這個專案沒有提供開發容器、Dockerfile 或貢獻指南,環境需要你自己搭建:先看它的 README,通用步驟見我們的新手貢獻指南。

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。

TanStack/oracles 的其他 Issue

查看 TanStack/oracles 的全部 Issue

相似的 Issue

更多 JavaScript 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。