Conversation
Split assert_test.sh (79 tests) into 4 focused files Split release_test.sh (71 tests) into 5 domain files Split coverage_test.sh (65 tests) into 4 functional files Improves parallel execution by distributing 215 tests across more workers.
Remove assert_test.sh, release_test.sh, and coverage_test.sh after splitting. Tests are now distributed across 13 focused files for better parallel load balancing.
Replace tail subprocess with parameter expansion. Eliminates 750+ process spawns during aggregation. Use (( )) arithmetic for counter updates (Bash 3.0 compatible).
…ribution Split bashunit_assert_subcommand_test.sh (20 tests) into 3 files: - bashunit_assert_basic_test.sh (8 tests): basic assertions and help - bashunit_assert_errors_test.sh (6 tests): error cases and deprecation - bashunit_assert_multi_test.sh (6 tests): multi-assertion mode Split bashunit_direct_fn_call_test.sh (19 tests) into 2 files: - bashunit_direct_fn_call_basic_test.sh (11 tests): basic assertions - bashunit_direct_fn_call_advanced_test.sh (8 tests): exit codes and failures Split bashunit_lifecycle_output_test.sh (12 tests) into 2 files: - bashunit_lifecycle_verbose_test.sh (2 tests): verbose output behavior - bashunit_lifecycle_visibility_test.sh (4 tests): hook visibility modes Improves parallel execution by distributing 51 acceptance tests across more workers.
…ance Enable parallel execution of individual test functions within files, not just parallel file execution. This significantly improves CPU utilization and reduces test execution time. Implementation: - Modified call_test_functions to run test functions in parallel - Added opt-out mechanism via `# bashunit: no-parallel-tests` marker - Tests within a file now execute concurrently using background jobs - Result aggregation already supports per-test result files Opt-out marker added to tests with shared state: - install_test.sh (shares ./lib and ./deps directories) - bashunit_report_html_test.sh (shares HTML output files) - bashunit_upgrade_test.sh (shares tmp/ directory) Performance improvement: - Parallel execution: 31.71s (down from ~40s, ~21% faster) - CPU utilization: 702% (up from ~480%) - Better parallelization across all available cores All 750 tests passing, all 1136 assertions correct.
Replace cp + rm with tar pipe for sandbox directory copying. This provides significant performance improvements by: - Excluding directories during copy (not after) - Using tar's optimized bulk operations - Eliminating separate rm -rf cleanup Performance impact: - Sandbox tests: 50s+ → 2.45s (~20x faster) - Full parallel suite: 31.71s → 26.56s (16% faster) - Overall from baseline: 42.94s → 26.56s (38% faster) Excluded directories: .git, .release-state, node_modules, .tasks, tmp
Contributor
✅ Contributor ReportUser: @Chemaclass
Contributor Report evaluates based on public GitHub activity. Analysis period: 2025-02-13 to 2026-02-13 |
Changes for Bash 3.0+ compliance:
- Replace [[ ]] with [ ] test operators in runner.sh
- Replace == with = for string comparisons
- Add fallback from tar to cp+rm for portability
- Fix indentation to pass EditorConfig linting
The tar pipe optimization now gracefully falls back to the traditional
cp + rm approach on systems without tar or with limited tar support,
ensuring maximum portability while maintaining performance gains on
systems with modern tar implementations.
Add no-parallel-tests marker to tests that have race conditions when run with test-level parallelism in strict mode: - parallel_test.sh: Tests parallel functionality, needs sequential execution to avoid interference with its own parallel temp directory management - bashunit_script_temp_file_cleanup_test.sh: Has data provider that runs bashunit internally, parallel execution causes tests to interfere with each other's temp file cleanup checks These tests were causing intermittent failures in strict parallel mode due to shared state and global temp directory access. Fixes race conditions while maintaining performance for other tests.
Replace [[ -d ]] && rm with if [ -d ]; then rm; fi pattern in tear_down function to ensure proper behavior in strict mode (set -euo pipefail). The [[ ]] && pattern can cause issues in strict mode when the test returns false, potentially causing the tear_down to exit early and prevent proper cleanup and variable restoration. This ensures Bash 3.0+ compatibility and stable behavior in all modes: - Normal mode: ✓ - Parallel mode: ✓ - Strict mode: ✓ - Parallel + Strict mode: ✓
Disable errexit temporarily around tar pipe to allow proper fallback to cp+rm when tar fails. In strict mode (set -euo pipefail), pipe failures would exit the function before the fallback could execute. Changes: - Wrap tar command with set +e / set -e - Capture tar exit status explicitly - Check status and fallback to cp if needed This fixes intermittent sandbox test failures in strict parallel mode while maintaining the performance benefit of tar on systems where it works.
Replace process substitution with command substitution and positional parameters.
Chemaclass
force-pushed
the
feat/improve-parallel-runner
branch
from
4acc337 to
cda6633
Compare
The disown && kill command could fail if the spinner process already finished, causing non-zero exit code in strict mode. Split into separate commands with || true to ensure they never fail.
The (( var += val )) syntax returns exit code 1 when result is 0, causing script to exit in strict mode with -e flag. Use var=$((var + val)) instead which never fails.
Add CHANGELOG entries for: - 30-40% performance improvement in parallel mode - Test-level parallelism feature - Opt-out directive for files with shared state - Strict mode compatibility fixes Add documentation for the no-parallel-tests directive in command-line guide.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
📚 Description
Improve parallel test execution by implementing test-level parallelism — individual test functions within a file now run concurrently when
--parallelis enabled. Previously, only file-level parallelism was supported. Also splits large test files for better parallel distribution and optimizes sandbox creation in the release script.🔖 Changes
&+wait)# bashunit: no-parallel-testsheader commenttailwith pure Bash parameter expansion and use(( ))arithmetic for countingassert_test.sh,coverage_test.sh,release_test.sh, and acceptance test files into smaller, focused files for better parallel distributioncp -r+rm -rfwith atarpipe that excludes directories upfront, avoiding unnecessary copying✅ To-do list
CHANGELOG.mdto reflect the new feature or fix🖼️ Demo
BEFORE
AFTER