Skip to content

Releases

v0.1.3

Choose a tag to compare

@github-actions github-actions released this 11 Nov 21:30

What's Changed

New Contributors

Full Changelog: v0.1.2...v0.1.3

v0.1.2

Choose a tag to compare

@qa-gary-parker qa-gary-parker released this 15 Oct 20:37

Better plans for suites with setup projects. Upgrade if your projects depend on setup or teardown projects and each shard runs more than one worker.

  • Setup projects that Playwright starts in the same phase now count as running side by side on the shard's workers. A teardown counts after the tests. Before, every setup was added one after another, so shards that need several setups were over-estimated.
  • The two best plans get a second pass that counts setup when it moves groups between shards. A costly setup family no longer spreads over every shard when one shard could take it.
  • Playwright's own split is scored as a candidate, so a plan is never predicted to be slower than the built-in split.
  • Benchmark, two projects with their own setup: 2 shards x 2 workers went from 3% slower than built-in to 8% faster; 4 shards x 2 workers from 13% to 20% faster.
  • Planning takes longer only when setup projects exist: about 1.35s for 10,000 groups on 4 shards with 8 workers.

v0.1.1

Choose a tag to compare

@qa-gary-parker qa-gary-parker released this 15 Oct 18:33

Fixes from a full review of 0.1.0. Upgrade if you shard suites of about 4,000 groups or more, use setup projects, or run the GitHub Actions.

  • Plans large suites again. 0.1.0 overflowed the stack at about 4,000 groups and failed every shard.
  • Better plans. A setup-aware pass helps suites whose projects need different setup projects. Swaps between shards help when each shard runs several workers.
  • Fewer false failures. Strict mode no longer fails a shard with no tests to split. --list and merge-reports runs no longer write run records. check names the tests behind each problem.
  • actions/restore: a pull request from a fork uses the built-in split instead of failing. Strict mode is reset per suite.
  • actions/finish: merges into the newest cached timings, so overlapping runs keep each other's durations. It saves only after update succeeds. The example installs the project, so finish runs the same CLI version as the reporter.

v0.1.0

Choose a tag to compare

@qa-gary-parker qa-gary-parker released this 15 Oct 14:57

First release.

  • A reporter that balances --shard=k/n by recorded duration through Playwright's preprocess hook (1.62+). CI needs no planning job.
  • Keeps Playwright's groups whole: non-parallel files, serial describes, and tests that share a lock. Counts setup projects and workers per shard.
  • CLI: update, simulate (shard-count advice), plan (--test-list files for Playwright 1.56+), and check (every test ran exactly once, plus a job summary).
  • GitHub Actions: actions/restore pins one copy of the timings per workflow run. actions/finish checks the run, then learns the new durations.