shell-escape.test.ts: Windows PowerShell child times out in the cmd-Session powershell test despite the warm-up

オープン
  • 何か問題がありますか？
#1,565 コメント 1 件 リアクション 0 件 担当者 0 名 GitHub で見る

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
4/5
見積もり時間
1〜2日
初心者へのやさしさ
42/100
issue の種類
バグ
明瞭さ
おおむね明確
活発さ
活発
技術スタック
github-actions, powershell, typescript
領域
ci-cd, testing

調査の方向性

Start with lib/src/lib/shell-escape.test.ts around lines 181–182 and inspect the warm-up added in #1532 and the failure context in #1564. On a Windows runner, time the listed PowerShell script pieces for each Session kind to locate the remaining cost. Done means identifying and addressing the cmd-kind delay, then confirming the test completes reliably without raising the per-shell timeout.

索引モデルが issue の本文から書いたものです。

説明

tend-ci-cause

shell-escape.test.ts › reads a cmd Session's paths back literally in powershell timed out again on main after the ConvertTo-Json warm-up from #1532 landed. In run 41778763200 (Standalone Platform Check (windows-latest), head 6c252bf967ba), argvThrough's spawnSync of powershell.exe returned ETIMEDOUT at its 10s bound (shell-escape.test.ts:181-182). The next run on main passed.

The warm-up ran and reported a warm shell. Its first start took 22,940 ms, against 2.4–4.4s on the other post-#1532 runs below, and its second took 437 ms, so the step exited at its sub-second threshold. The Windows PowerShell 5.1 children around the failing test were fast:

Test (in run order) Time
runs native PowerShell startup code and exits 524 ms
passes accepted paths through dor.cmd from powershell 480 ms
reads a cmd Session's paths back literally in powershell 10,048 ms, ETIMEDOUT
reads a powershell Session's paths back literally in powershell 572 ms

A cold process start doesn't explain this, and neither does a first module load the warm-up skips: the 5.1 children on either side of the failing test, the later one also running ConvertTo-Json, took about half a second.

On the eight main runs since #1532 whose Windows job passed (41578300835 through 41711627101), the failing test took 697–799 ms each time, while reads a powershell Session's paths back literally in powershell took 303–389 ms. #1532 cut it from the 1.2–3.1s seen before, but some first-use cost the warm-up doesn't reach is still left. On a runner as slow as this one, that cost alone doesn't obviously add up to 10s (≈400 ms × 5–10 ≈ 2–4s). So either the remaining cost grows faster than the runner's slowness, or this run hit a separate stall. This run doesn't settle which.

In #1464, nedtwigg ruled out raising the per-shell bounds ("raising either bound again would only move the line"), so that isn't proposed here.

Possible next diagnostics (need a Windows runner)
  • Find what the cmd-kind 5.1 child pays that the warm-up and the powershell-kind child don't. Its script is the same argv function; only the typed arguments differ (the cmd-kind escapes of WINDOWS_LITERAL and WINDOWS_REFUSED). Timing each piece of that script in the warm-up step ([Console]::OutputEncoding, the function definition, ConvertTo-Json -Compress -InputObject @(...) over those strings) on a PR's Windows job would show which one carries the ~400 ms.
  • If the cost tracks script content rather than process start (for example, a first scan of a new script block), a warm-up that runs the exact argvThrough script for each Session kind would pay it before the tests. The check: on that PR's job, the cmd-kind test should drop to about the powershell-kind test's ~300 ms.

Cause of a failure tracked in #1564.

主要言語
TypeScript
スター
5
フォーク
1
平均マージ
11時間 41分
マージ済み PR（30日）
432

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

  1. 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。
  2. 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。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

diffplug/dormouse のほかの issue

diffplug/dormouse の issue をすべて見る

似ている issue

TypeScript の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。