[robobun]

Problem

• A Bun.file() read that knows the file size skips the last read() (0 bytes at EOF). Today only a read after .size or exists() knows it.
• A read after .lastModified or Response.clone() has a statted store and no size on the blob. It pays that read(): 2.04 per text() of a 64 KiB file, not 1.03.
• Bun.file: do not use a cached stat size as the byte budget of a read or a copy #43910 takes st_size off the blob (a procfs file reports 0 and has content). Every stat-first read would then pay it.

Fix

• ReadFile, ReadFileUV and FileReader stop at the size from the fstat they run at open. On Windows FileReader asks the handle (GetFileSizeEx).
• This applies only when the store was statted before the read. A size of 0 is never a bound.
• Verified: test/js/bun/util/bun-file-read.test.ts counts read syscalls (/proc/self/io): 2 on main, 1 here. More in Notes.

Background

• Every Blob over one Bun.file() shares a store that caches one stat. Only resolve_file_stat writes seekable, so seekable.is_some() means "statted".
• Considered the cached size as the limit when it is not 0. It goes stale: exists(), a write, then text() stops at the old length (BunFile .text() does not return correct content after BunFile.write() #23902). A fresh fstat costs the same.

Downsides

• A file that grows between the fstat and the last read() yields its fstat-time bytes for these reads. A read after .size already stops at an older size.
• Windows: one GetFileSizeEx per stream of a statted store. POSIX: no new syscall.

Notes

read syscalls per operation, 64 KiB file, /proc/self/io (syscr) over 300 iterations, debug builds, Linux x64. Syscall counts do not depend on the build profile.

operation	main	this PR
fresh text()	2.03	2.04
.size, then text()	1.03	1.03
exists(), then text()	1.03	1.03
.lastModified, then text()	2.04	1.04
fresh stream()	2.10	2.10
.size, then stream()	1.09	1.09
.lastModified, then stream()	2.10	1.09
new Response(file).body	1.10	1.09

Windows x64, read operations per operation from Win32_Process.ReadOperationCount, same file, 300 iterations, debug builds: .lastModified then text() 2.01 to 1.01, .lastModified then stream() 2.01 to 1.01. Fresh reads stay at 2.01 and stat-first reads stay at 1.01.

The gate is file_store.seekable.is_some() on the File that the reader captured when it was created. A fresh Bun.file(p).text() keeps its behaviour: it reads until EOF, and it returns bytes that were appended after its fstat.

A file whose fstat under-reports a size above 0 is cut at that size for these reads, as .size then text() already is on main. A size of 0 (procfs, sysfs with no content length) is never a bound.

The plan for this stack called this PR a refactor with no behaviour change. That is not exact, because of the first Downsides line, so this PR has a test and does not claim it.

Suites run with a debug build, Linux: blob, body, body-clone, wasm-streaming, structured-clone-blob-file (1044 pass), streams, fetch-file-upload, bun-file-fd-read, empty-file (638 pass). fetch.test.ts, bun-file-exists, bun-file, bun-file-read, bun-serve-file, FormData: the same 21 failures by name as main (they need IPv6, the public internet, or a non-root user). Windows: 1719 pass in 13 suites. bun run rust:check-all passes on 12 targets.

[robobun]

Status: draft. This is the second PR of a stack: #43897 (refactor, green), this PR, then #43910 (the fix).

What it changes, measured on Linux x64 with /proc/self/io (read syscalls per operation, 64 KiB file, 300 iterations):

const f = Bun.file(p);
f.lastModified;          // stats the store, the blob keeps no size
await f.text();          // main: 2.04 reads. here: 1.04

Fresh reads (2.03) and reads after .size or exists() (1.03) do not move. On Windows x64 the same case goes from 2 uv_fs_read calls to 1.

The new test in test/js/bun/util/bun-file-read.test.ts fails with src/ at main (2 reads per operation) and passes here (1).

It stays a draft until the review of the whole stack is done.

[robobun]

Closing: this change is now part of #43910.

The reader bound moves the byte budget of a read from the cached stat to the fstat of the reader. #43910 removes the cached budget. The two are one change, and this half changes behaviour, so it is not groundwork that can merge alone.

#43910 has the same reader code with one difference: ReadFile and ReadFileUV share one helper. The measurements from this PR are in the Notes of #43910, from release builds. The branch stays, so this PR can reopen if the split is preferred.