v3.2.0

3.2.0 - 2026-07-18

• TAG: v3.2.0
• COVERAGE: 96.88% -- 869/897 lines in 10 files
• BRANCH COVERAGE: 84.44% -- 217/257 branches in 10 files
• 34.69% documented

Added

• --count is now accepted as an alias for -n, matching the
  parallel_tests process-count option while keeping the existing
  -w / --workers aliases.
• Documented the supported parallel_tests option subset and clarified that
  unsupported flags are not blindly passed through after --.
• Added --worker-output MODE and TURBO_TESTS2_WORKER_OUTPUT support for
  selecting raw worker stdout/stderr handling: warnings, stream, buffered,
  or quiet.
• --group-by runtime|filesize|found is now parsed explicitly and mapped to
  parallel_tests grouping modes, including when supplied after the
  compatibility -- separator.
• --allowed-missing PERCENT and --unknown-runtime SECONDS are now parsed
  explicitly and mapped to parallel_tests runtime grouping tuning options,
  including when supplied after the compatibility -- separator.

Changed

• Pull request engine workflows now keep JRuby and TruffleRuby coverage
  available via branch opt-in prefixes: jruby/* runs JRuby workflows,
  truffleruby/* runs TruffleRuby workflows, and engines/* runs all engine
  workflows. Other pull request branches continue to run MRI checks without the
  alternate engine jobs.
• The delayed documentation formatter spec now lives under fixtures so the
  normal suite does not pay its fixed load-time delay.

Fixed

• Successful non-verbose runs now surface warning and deprecation lines emitted
  as raw worker stdout/stderr without dumping all buffered worker output.
• heads.yml and dep-heads.yml now gate JRuby and TruffleRuby PR coverage
  at the job level, so non-matching PR branches skip those engine jobs before a
  runner is provisioned.
• Direct turbo_tests2 runs now refresh the default
  tmp/turbo_rspec_runtime.log timing log, and --runtime-log writes the
  requested timing log even when the run groups by filesize.
• --only-group GROUP_INDEX[,GROUP_INDEX] is now parsed explicitly and runs
  only the selected 1-based parallel_tests group index(es), using filesize
  grouping by default for selected-group reruns.
• --pattern PATTERN is now parsed explicitly and mapped to parallel_tests
  file discovery instead of being treated as a spec file when supplied after
  --.
• --exclude-pattern PATTERN is now parsed explicitly and mapped to
  parallel_tests file discovery instead of being treated as a spec file when
  supplied after --.

Official Discord 👉️

Many paths lead to being a sponsor or a backer of this project. Are you on such a path?