nodejs/build

Measure flakiness of new tests

Open

#3,541 opened on Oct 25, 2023

View on GitHubContributor guide
 (4 comments) (0 reactions) (0 assignees)Jinja (180 forks)auto 404
enhancementhelp wantedquestion

Repository metrics

Stars
 (537 stars)
PR merge metrics
 (PR metrics pending)

Description

As discussed in https://github.com/nodejs/TSC/issues/1457, could we somehow have a way for CI to measure the flakiness of new tests before they land? Something like:

  • For every PR, identify tests that are added by the PR (probably tests that run in the PR’s branch that didn’t run for main).
  • Run measure-flakiness on them.
  • Fail CI unless the new tests pass the flakiness cutoff, on all platforms.

This obviously won’t help for existing flaky tests, but I would think that it should prevent most new flaky tests from landing on main; and it would highly motivate contributors to improve their tests, because their PRs would be blocked from landing until they did so.

It also wouldn’t help if a test becomes flaky because of changes to the API that it tests after the test has landed. But still, I think this is better than the status quo.

Related: https://github.com/nodejs/build/pull/3056 cc @nodejs/tsc

Contributor guide

Research direction
Investigate how to identify new tests in a PR and run them multiple times to measure flakiness. Look into CI workflow modifications and possibly using tools like test retries or flaky test detection frameworks.
Tech stack
None
Domain
ci cdtesting
Issue type
Feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
Half day
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
CI/CDTesting
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50