test_runner: avoid recompiling coverage globs for every file

[Han5991]

This speeds up the --experimental-test-coverage report phase by removing redundant glob work in shouldSkipFileCoverage(). Refs: #55103 (the report-generation half of that issue).

Background

When building the coverage report, shouldSkipFileCoverage(url) runs once per covered file to decide whether the file belongs in the report (e.g. test files are excluded by default). For each exclude/include glob it calls matchGlobPattern(path, pattern), and matchGlobPattern() constructs a fresh Minimatch — a full glob parse + regexp compile — on every call (lib/internal/fs/glob.js).

The exclude/include globs don't change during a run, so the same patterns are recompiled once per file. cpu-profiling a coverage run shows this glob compilation (shouldSkipFileCoverage → matchGlobPattern → minimatch make/parse/regexp compile) as the dominant cost of the report path.

Changes

Two commits:

1. test_runner: avoid recompiling coverage globs for every file — the main change.
The globs are fixed for the lifetime of a TestCoverage instance, so compile each to a matcher once and reuse it for every file instead of recompiling per call.

• internal/fs/glob: export the existing createMatcher() helper (already used internally by Glob to precompile patterns).
• test_runner/coverage: build #excludeMatchers / #includeMatchers once, then match with matcher.match(path).

No behavior change — same patterns, same options, same match results.

2. test_runner: cache shouldSkipFileCoverage result per URL — complementary.
Memoizes the per-URL skip decision. Under isolation: 'process' the same script URL is reported by every worker, so this dedups the computes, while commit 1 makes each compute cheap. The result depends only on the URL plus the instance's fixed options (cwd, exclude/include globs), so it is safe to cache for the instance lifetime.

Measurements

Synthetic project: 200 test files importing 80 shared src modules, --experimental-test-coverage, default exclude glob, --test-isolation=none, cpu-prof of the report-generating process. Median of 3 runs; before/after measured on the same build (the glob commit reverted via git stash + rebuild):

metric	before	after
shouldSkipFileCoverage (inclusive, incl. minimatch)	~117 ms	~11 ms
minimatch self-time (glob compile)	~111 ms	~11 ms

The saving is independent of test weight and scales with files × globs, so it grows on large suites (which are the ones that notice the report lag) and is negligible on small ones.

Correctness

Existing coverage tests pass, including test/parallel/test-runner-coverage-default-exclusion (which exercises the default exclude behavior) and the rest of test-runner-coverage* / test-runner-run-coverage.

[nodejs-github-bot]

Review requested:

• @nodejs/test_runner

[mcollina]

lgtm

[codecov]

Codecov Report

❌ Patch coverage is 93.10345% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 90.33%. Comparing base (8b88e2c) to head (226a920).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
lib/internal/test_runner/coverage.js	92.85%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #63675      +/-   ##
==========================================
- Coverage   91.95%   90.33%   -1.63%     
==========================================
  Files         379      732     +353     
  Lines      166486   236465   +69979     
  Branches    25452    44527   +19075     
==========================================
+ Hits       153099   213620   +60521     
- Misses      13094    14548    +1454     
- Partials      293     8297    +8004     

Files with missing lines	Coverage Δ
lib/internal/fs/glob.js	91.49% <100.00%> (+3.37%)	⬆️
lib/internal/test_runner/coverage.js	62.90% <92.85%> (+0.22%)	⬆️

... and 479 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.