[robobun]

What does this PR do?

fs.readv(fd, []), fs.readvSync(fd, []), and FileHandle#readv([]) now throw EINVAL instead of returning bytesRead: 0, matching Node.js.

Reproduction

import * as fs from "node:fs";
const fd = fs.openSync("/etc/hosts", "r");
try { console.log(fs.readvSync(fd, [])); } catch (e) { console.log(e.code, e.syscall); }

Node v26.3.0: EINVAL read
Bun before: 0
Bun after: EINVAL read

Cause

Node routes readv through uv_fs_read, which has an explicit nbufs == 0 -> UV_EINVAL guard. Node only short-circuits the empty-array case for writev (in JS, before reaching libuv), so writev(fd, []) returns 0 while readv(fd, []) throws. Bun's readv() in src/runtime/node/node_fs.rs was short-circuiting both to 0.

Fix

Change the readv empty-buffers short-circuit to return EINVAL with syscall "read", matching Node's error object exactly ({code: "EINVAL", errno: -22, syscall: "read"}). The Windows libuv async path gets the same guard; without it, uv_fs_read would return UV_EINVAL synchronously rather than via the callback, leaving the request unresolved.

writev(fd, []) is unchanged and still returns 0.

How did you verify your code works?

New tests in test/js/node/fs/fs.test.ts cover all three forms (sync, callback, FileHandle) plus the positional variant, and assert both the readv -> EINVAL and writev -> 0 sides. All fail on current main and pass with the fix. bun run rust:check-all passes for all targets.

---

no test proof · iteration 0 · Platform-specific test(s) that do not run on this machine. Deferring to CI, which covers all platforms: test/js/node/fs/fs.test.ts

[robobun]

^{Updated 1:10 PM PT - Jul 16th, 2026}

⏳ @robobun, your commit 254f2fc is still building in Build #74025, but has 1 failures so far (All Failures):

• test/js/node/test/parallel/test-worker-message-port-transfer-terminate.js - SIGABRT on 🐧 13 x64-asan

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Pro

Run ID: c80eec81-f412-4fda-9b7c-111479f85a53

📥 Commits

Reviewing files that changed from the base of the PR and between 6ad467f and 254f2fc.

📒 Files selected for processing (2)

• src/runtime/node/node_fs.rs
• test/js/node/fs/fs.test.ts

---

Walkthrough

Changes

Empty-buffer readv behavior

Layer / File(s)	Summary
Readv validation and API coverage src/runtime/node/node_fs.rs, test/js/node/fs/fs.test.ts	Synchronous and Windows asynchronous readv now return EINVAL with syscall read for empty buffers; corresponding readv and writev APIs are covered by tests.

Suggested reviewers: jarred-sumner

🚥 Pre-merge checks | ✅ 4

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely summarizes the main change: empty fs.readv buffers now throw EINVAL.
Description check	✅ Passed	The description matches the template and includes the change summary plus verification details with reproduction and tests.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

---

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

[claude]

LGTM — small Node.js compat fix matching libuv's uv_fs_read(nbufs==0) → UV_EINVAL behavior.

Checked that the Windows async guard mirrors the existing Writev empty-buffers pattern at node_fs.rs:889-903 (same core::ptr::write + enqueue_task sequence), and that the sync path's sys::Error construction matches neighboring error sites. Verified writev([]) remains a 0-return on both the sync and Windows-async paths. Tests cover sync/callback/FileHandle plus the positional variant, use tempDir with proper cleanup, and pin the unchanged writev contract.

Extended reasoning...

Overview

Two-file change: src/runtime/node/node_fs.rs swaps the empty-buffers short-circuit in readv from Ok(bytes_read: 0) to Err(EINVAL, syscall=read), and adds a matching guard to the Windows libuv async Readv arm so uv_fs_read is never called with nbufs==0 (which would return synchronously and never fire the callback). test/js/node/fs/fs.test.ts gains a describe block covering sync, callback, and FileHandle forms for both readv (now EINVAL) and writev (still 0).

Security risks

None. No new input parsing, no path handling, no allocation sizing — this replaces a success return with an error return on an already-validated empty array.

Level of scrutiny

Low. This is a targeted Node.js compatibility fix with a well-documented upstream rationale (libuv's explicit nbufs==0 → UV_EINVAL guard, which Node surfaces for readv but short-circuits in JS for writev). The Windows async change is a byte-for-byte copy of the existing Writev empty-buffers handling immediately below it (lines 889-903), just writing Err instead of Ok. The sync change is a one-expression swap using the same sys::Error { errno, syscall, fd, ..Default::default() } shape as the neighboring preadv_inner error path.

Other factors

Tests follow harness conventions (tempDir, using/await using, try/finally around fd), assert exact error shape (code + syscall), and cover the positional overload and the negative contract (writev([]) unchanged). The bug hunting system found no issues. rust:check-all reportedly passes for all targets.

[robobun]

Closing as part of a cleanup of stale pull requests. This PR has had no new commits since 2026-07-16, it conflicts with main, and its last CI run failed. This is not a judgment on the fix itself. If the problem still reproduces on a current build, reopen this PR after a rebase or open a new one against main.