[robobun]

Summary

• Fix fs.stat truncating sub-millisecond precision from mtimeMs, atimeMs, ctimeMs, and birthtimeMs properties
• The non-BigInt path in toTimeMS() was using bun.timespec.ms() which performs integer division, losing the nanosecond remainder
• Switch to floating-point arithmetic so fractional milliseconds are preserved, matching Node.js behavior

Before

$ bun mtime-test.js
mtimeMs: 1773248895434

After

$ bun mtime-test.js
mtimeMs: 1773248895434.0544

Test plan

• Added regression test test/regression/issue/28017.test.ts that verifies *Ms properties have fractional components
• Test fails with system bun (USE_SYSTEM_BUN=1) and passes with debug build
• Existing stat tests (fs-stats-truncate.test.ts) still pass

Closes #28017
Closes #21943
🤖 Generated with Claude Code

[robobun]

^{Updated 12:01 PM PT - Mar 11th, 2026}

❌ Your commit db774fb6 has 2 failures in Build #39311 (All Failures):

• test/js/bun/util/v8-heap-snapshot.test.ts - SIGKILL on 🐧 25.04 x64-baseline
• test/integration/next-pages/test/dev-server.test.ts - code 1 on 🍎 14 aarch64

---

🧪   To try this PR locally:

bunx bun-pr 28018

That installs a local version of the PR into your bun-28018 executable, so you can run:

bun-28018 --bun

[coderabbitai]

Walkthrough

Replaces truncating millisecond computation with a floating-point calculation in Stat.toTimeMS to preserve sub-millisecond precision; adds a regression test that verifies fractional atimeMs and mtimeMs for both synchronous and asynchronous stat results.

Changes

Cohort / File(s)	Summary
Stat time conversion & regression test src/bun.js/node/Stat.zig, test/regression/issue/28017.test.ts	Replaces integer/truncating ms calculation in toTimeMS with f64 computation (sec * 1000 + nsec / 1_000_000) to preserve fractional milliseconds and match Node.js; adds a test asserting non-integer atimeMs/mtimeMs for both sync and async stat.

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title 'fix(fs): preserve sub-millisecond precision in stat *Ms properties' clearly and specifically describes the main change in the PR.
Linked Issues check	✅ Passed	The code changes directly address issue #28017 by implementing floating-point arithmetic in toTimeMS to preserve sub-millisecond precision in fs.stat *Ms properties, matching Node.js behavior.
Out of Scope Changes check	✅ Passed	All changes are directly scoped to the stated objective: the fix modifies the time-conversion logic and adds a regression test, with no extraneous modifications.
Description check	✅ Passed	The PR description comprehensively covers the issue, solution, before/after examples, and test plan with checkboxes.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@test/regression/issue/28017.test.ts`:
- Around line 5-24: Replace the nondeterministic use of import.meta.path with a
deterministic temp file created via harness.tempDir: create a tempDir, write a
test file into it, set known fractional atime/mtime using fs.utimes or
fs.utimesSync (so the timestamps include a fractional millisecond component you
control), then call statSync and await stat on that temp file and assert that
the returned mtimeMs/atimeMs (and other *Ms values if needed) exactly match the
expected fractional values; update the test named "fs.stat *Ms properties
preserve sub-millisecond precision" to remove reliance on import.meta.path and
use the tempDir-created file and controlled utimes instead.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: ad912789-03f3-45c5-9c9e-5df1056baa0b

📥 Commits

Reviewing files that changed from the base of the PR and between fbfd069 and 7b502c6.

📒 Files selected for processing (2)

• src/bun.js/node/Stat.zig
• test/regression/issue/28017.test.ts

[github-actions]

Found 1 issue this PR may fix:

1. fs.stat.birthtimeMs returns an integer instead of a float #21943 - fs.stat.birthtimeMs returns an integer instead of a float with sub-millisecond precision

If this is helpful, consider adding 'Fixes #21943' to the PR description to auto-close the issue on merge.

🤖 Generated with Claude Code

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@test/regression/issue/28017.test.ts`:
- Around line 26-29: The test currently uses
expect(s.atimeMs).toBeCloseTo(expectedMs, 0) and
expect(s.mtimeMs).toBeCloseTo(expectedMs, 0) which permits ±0.5ms; tighten this
to assert microsecond-level determinism by changing the matcher to use a higher
precision (e.g. toBeCloseTo(expectedMs, 3)) so s.atimeMs and s.mtimeMs are
compared to expectedMs to within 0.001ms (reference symbols: s.atimeMs,
s.mtimeMs, expectedMs, fractionalTime in the test).

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: 5319039a-c27a-4fa8-a488-32a39aff6bef

📥 Commits

Reviewing files that changed from the base of the PR and between 7b502c6 and 85ce1a5.

📒 Files selected for processing (1)

• test/regression/issue/28017.test.ts