Should a Mock inside a synctest test follow synctest's clock?

未关闭
  • 哪里不对？
#35 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

评估

难度
5/5
预计耗时
一周以上
新手友好度
8/100
Issue 类型
功能
描述清晰度
基本清楚
活跃度
活跃
技术栈
go
领域
testing-qa

调研方向

This is a design question, not a patch: the issue asks whether a *Mock inside a synctest test should follow synctest's fake clock or keep its own time. Read the mock clock implementation and the NewReal() clock in the quartz package, along with the prototype behaviour described in the issue (Advance, timers, ctx.Deadline). Done means a maintainer-agreed answer to the design question and the README wording that follows from it, which needs input from @spikecurtis first.

由索引模型根据 Issue 内容生成。

描述

Go 1.25 made testing/synctest stable. In a synctest test, the time package runs on a fake clock that jumps forward whenever every goroutine in the test is blocked. That overlaps with what quartz does, so the README needs to say when to use which (#34). Before writing that, we want to agree on one design question:

Should a *Mock created inside a synctest test follow synctest's fake clock, or keep its own time as it does today?

Today the two clocks are independent. Calling time.Sleep(time.Hour) inside the test does not move the mock, and mClock.Advance(time.Hour) does not move time.Now(). quartz.NewReal() already follows synctest's clock, because it calls the time package.

We prototyped a mock that follows synctest's clock: timers are real time timers, and Advance(d) is time.Sleep(d) followed by synctest.Wait(). Compared with today's *Mock:

  • Better: ctx.Deadline() agrees with time.Now(), and Advance can move past several timers at once, firing them in order. Today Advance stops at the next timer.
  • Worse, time moves without Advance: a test that forgot to call Advance and then waited for the code to start its next timer still passed, because synctest moved time forward by 10s on its own. Today's mock fails that test.
  • Worse, real I/O: a 5s timeout never fired during a real TCP read. Advance(5s) blocked until the read finished, because synctest does not move time while a goroutine waits on the network. Today's mock canceled the read after 67µs.

We also tried keeping the mock's own timers but sleeping the same amount on every Advance, so both clocks stay equal. They drift apart as soon as code in the test sleeps on its own. After a 1-minute time.Sleep in library code, the mock was 1 minute behind.

The one thing a synctest-following mock adds over NewReal() is traps.

Our recommendation: keep the *Mock independent. The README would say: use synctest with NewReal() when the whole test can run inside synctest, and use a *Mock when it does real I/O or needs to control exactly when time moves. If traps turn out to be wanted with synctest, a separate option is to add traps to the real clock rather than changing the mock.

@spikecurtis, does that match your view?

🤖 This issue was created with the help of Coder Agents.

主要语言
Go
星标
273
派生
9
30 天内没有已合并 PR

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

coder/quartz 的其他 Issue

查看 coder/quartz 的全部 Issue

相似的 Issue

更多 Go Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。