Add FileRoute for serving files #20198

Jarred-Sumner · 2025-06-05T03:04:55Z

Summary

implement FileRoute for serving files
support parsing Bun.file() or Response with file body
integrate FileRoute into AnyRoute
handle sendfile responses through uws.AnyResponse
fixes support Bun.file on static routes #17362

Testing

bun bd test internal/ban-words.test.ts (fails: missing WebKit)

https://chatgpt.com/codex/tasks/task_e_6840cce283088328bef19cdb97982d22

robobun · 2025-06-05T03:05:07Z

^{Updated 8:25 PM PT - Jun 10th, 2025}

❌ @dylan-conway, your commit e181163 has 3 failures in Build #18466:

test/js/node/http/node-http-backpressure.test.ts - SIGKILL on 🐧 20.04 aarch64
test/js/bun/spawn/spawn.test.ts - 1 failing on 🪟 2019 x64-baseline
test/js/bun/spawn/spawn.test.ts - 1 failing on 🪟 2019 x64
annotation error on 🍎 13 aarch64 - test-bun

🧪 To try this PR locally:

bunx bun-pr 20198

That installs a local version of the PR into your bun-20198 executable, so you can run:

bun-20198 --bun

RiskyMH · 2025-06-05T12:23:18Z

also don't forget to update types

bun/packages/bun-types/bun.d.ts

Line 3298 in f62940b

    
           type RouteValue<T extends string> = Response | false | RouteHandler<T> | RouteHandlerObject<T>;

TiBianMod · 2025-06-05T17:06:44Z

Good new 👍
@Jarred-Sumner, also related to routes #17363
will be nice addition

src/bun.js/api/server/FileRoute.zig

src/bun.js/bindings/WTF.zig

graphite-app · 2025-06-06T03:25:44Z

src/fs/stat_hash.zig

+value: u64 = 0,
+
+last_modified_u64: u64 = 0,
+last_modified_buffer: [32]u8 = undefined,


The field last_modified_buffer is initialized with undefined, which violates the project's rule against defaulting struct fields to undefined. This can lead to undefined behavior if the buffer is accessed before being properly initialized. Consider using a zero-initialized array (e.g., = [_]u8{0} ** 32) or ensure the buffer is always initialized before use. The ban-words test has been updated to allow one more occurrence of this pattern, but fixing the underlying issue would be preferable to incrementing the limit.

Suggested change

last_modified_buffer: [32]u8 = undefined,

last_modified_buffer: [32]u8 = [_]u8{0} ** 32,

Spotted by Diamond (based on custom rules)

Is this helpful? React 👍 or 👎 to let us know.

graphite-app · 2025-06-06T04:42:51Z

src/fs/stat_hash.zig

+    if (prev != this.value and bun.S.ISREG(@intCast(stat.mode))) {
+        const mtime_timespec = stat.mtime();
+        // Clamp negative values to 0 to avoid timestamp overflow issues on Windows
+        const mtime = bun.timespec{
+            .nsec = @intCast(@max(mtime_timespec.nsec, 0)),
+            .sec = @intCast(@max(mtime_timespec.sec, 0)),
+        };
+        if (mtime.ms() > 0) {
+            this.last_modified_buffer_len = @intCast(bun.JSC.wtf.writeHTTPDate(&this.last_modified_buffer, mtime.msUnsigned()).len);
+            this.last_modified_u64 = mtime.msUnsigned();
+        } else {
+            this.last_modified_buffer_len = 0;
+            this.last_modified_u64 = 0;
+        }
+    } else if (!bun.S.ISREG(@intCast(stat.mode))) {
+        this.last_modified_buffer_len = 0;
+        this.last_modified_u64 = 0;
+    }


The current implementation only updates last_modified values when both the hash changes AND the file is a regular file. This could miss setting these values on the first hash calculation when prev == 0.

Consider restructuring to:

if (bun.S.ISREG(@intCast(stat.mode))) { const mtime_timespec = stat.mtime(); // Clamp negative values to 0 to avoid timestamp overflow issues on Windows const mtime = bun.timespec{ .nsec = @intCast(@max(mtime_timespec.nsec, 0)), .sec = @intCast(@max(mtime_timespec.sec, 0)), }; if (mtime.ms() > 0) { this.last_modified_buffer_len = @intCast(bun.JSC.wtf.writeHTTPDate(&this.last_modified_buffer, mtime.msUnsigned()).len); this.last_modified_u64 = mtime.msUnsigned(); } else { this.last_modified_buffer_len = 0; this.last_modified_u64 = 0; } } else { this.last_modified_buffer_len = 0; this.last_modified_u64 = 0; }

This ensures the values are always set for regular files, with the hash check only used as an optimization if needed.

Suggested change

if (prev != this.value and bun.S.ISREG(@intCast(stat.mode))) {

const mtime_timespec = stat.mtime();

// Clamp negative values to 0 to avoid timestamp overflow issues on Windows

const mtime = bun.timespec{

.nsec = @intCast(@max(mtime_timespec.nsec, 0)),

.sec = @intCast(@max(mtime_timespec.sec, 0)),

};

if (mtime.ms() > 0) {

this.last_modified_buffer_len = @intCast(bun.JSC.wtf.writeHTTPDate(&this.last_modified_buffer, mtime.msUnsigned()).len);

this.last_modified_u64 = mtime.msUnsigned();

} else {

this.last_modified_buffer_len = 0;

this.last_modified_u64 = 0;

}

} else if (!bun.S.ISREG(@intCast(stat.mode))) {

this.last_modified_buffer_len = 0;

this.last_modified_u64 = 0;

}

if (bun.S.ISREG(@intCast(stat.mode))) {

const mtime_timespec = stat.mtime();

// Clamp negative values to 0 to avoid timestamp overflow issues on Windows

const mtime = bun.timespec{

.nsec = @intCast(@max(mtime_timespec.nsec, 0)),

.sec = @intCast(@max(mtime_timespec.sec, 0)),

};

if (mtime.ms() > 0) {

this.last_modified_buffer_len = @intCast(bun.JSC.wtf.writeHTTPDate(&this.last_modified_buffer, mtime.msUnsigned()).len);

this.last_modified_u64 = mtime.msUnsigned();

} else {

this.last_modified_buffer_len = 0;

this.last_modified_u64 = 0;

}

} else {

this.last_modified_buffer_len = 0;

this.last_modified_u64 = 0;

}

Spotted by Diamond

Is this helpful? React 👍 or 👎 to let us know.

src/bun.js/api/server/FileRoute.zig

graphite-app · 2025-06-06T04:45:20Z

src/bun.js/bindings/webcore/HTTPParsers.cpp

+    return snprintf(buffer, length, "%.3s, %.2u %.3s %.4u %.2u:%.2u:%.2u GMT",
+        wday_name[tstruct.tm_wday],
+        tstruct.tm_mday % 99,
+        mon_name[tstruct.tm_mon],
+        (1900 + tstruct.tm_year) % 9999,
+        tstruct.tm_hour % 99,
+        tstruct.tm_min % 99,
+        tstruct.tm_sec % 99);


The snprintf format specifiers use %.2u which expects values between 0-99, but the modulo operations use % 99 which allows values up to 98. For consistency and to prevent potential issues, either:

Change the modulo operations to match the format width: tstruct.tm_mday % 100 for %.2u, or

Use %02u format specifiers and remove the modulo operations entirely since tm struct fields are already properly bounded

The second approach is preferred as it's cleaner and relies on the standard guarantees of the tm struct.

Spotted by Diamond

Is this helpful? React 👍 or 👎 to let us know.

graphite-app · 2025-06-07T02:10:09Z

src/bun.js/api/server/FileRoute.zig

+            response.body.value = .{ .Blob = blob.dupe() };
+            const headers = Headers.from(response.init.headers, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch bun.outOfMemory();


There's a potential memory leak in this code. The blob is duplicated with blob.dupe() and assigned to response.body.value before Headers.from() is called. If Headers.from() fails with any error other than out-of-memory, the duplicated blob won't be properly cleaned up.

Consider either:

Moving the blob duplication after the headers creation, or

Using a temporary variable and ensuring proper cleanup in case of failure

// Option 1: Move duplication after headers creation const headers = Headers.from(response.init.headers, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch bun.outOfMemory(); response.body.value = .{ .Blob = blob.dupe() }; // Option 2: Use temporary and ensure cleanup var duped_blob = blob.dupe(); const headers = Headers.from(response.init.headers, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch |err| { duped_blob.deinit(); return bun.outOfMemory(); }; response.body.value = .{ .Blob = duped_blob };

Suggested change

response.body.value = .{ .Blob = blob.dupe() };

const headers = Headers.from(response.init.headers, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch bun.outOfMemory();

const headers = Headers.from(response.init.headers, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch bun.outOfMemory();

response.body.value = .{ .Blob = blob.dupe() };

Spotted by Diamond

Is this helpful? React 👍 or 👎 to let us know.

Co-authored-by: graphite-app[bot] <96075541+graphite-app[bot]@users.noreply.github.com>

graphite-app · 2025-06-07T12:26:40Z

src/bun.js/api/server/FileRoute.zig

+    if (fd_result == .err) {
+        req.setYield(true);
+        this.deref();
+        return;
+    }
+
+    const fd = fd_result.result.makeLibUVOwned() catch {
+        req.setYield(true);
+        this.deref();
+        return;
+    };


The code correctly handles errors from openA but has a potential file descriptor leak when makeLibUVOwned fails. In this case, the file descriptor obtained from fd_result.result should be closed before returning. Consider adding a cleanup step:

const fd = fd_result.result.makeLibUVOwned() catch { // Close the raw file descriptor to prevent leaks bun.sys.close(fd_result.result); req.setYield(true); this.deref(); return; };

This ensures proper resource cleanup in all error paths.

Suggested change

if (fd_result == .err) {

req.setYield(true);

this.deref();

return;

}

const fd = fd_result.result.makeLibUVOwned() catch {

req.setYield(true);

this.deref();

return;

};

if (fd_result == .err) {

req.setYield(true);

this.deref();

return;

}

const fd = fd_result.result.makeLibUVOwned() catch {

// Close the raw file descriptor to prevent leaks

bun.sys.close(fd_result.result);

req.setYield(true);

this.deref();

return;

};

Spotted by Diamond

Is this helpful? React 👍 or 👎 to let us know.

graphite-app · 2025-06-10T19:40:39Z

src/bun.js/api/server/FileRoute.zig

+pub fn initFromBlob(blob: Blob, opts: InitOptions) *FileRoute {
+    const headers = Headers.from(null, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch bun.outOfMemory();
+    return bun.new(FileRoute, .{
+        .ref_count = .init(),
+        .server = opts.server,
+        .blob = blob,
+        .headers = headers,
+        .status_code = opts.status_code,
+    });
+}


The has_last_modified_header and has_content_length_header fields are uninitialized in this constructor, which could lead to undefined behavior when these fields are accessed later. Consider adding these initializations to the struct:

return bun.new(FileRoute, .{ .ref_count = .init(), .server = opts.server, .blob = blob, .headers = headers, .status_code = opts.status_code, .has_last_modified_header = headers.get("last-modified") != null, .has_content_length_header = headers.get("content-length") != null, });

This ensures the fields are properly initialized with the correct values based on the headers.

Suggested change

pub fn initFromBlob(blob: Blob, opts: InitOptions) *FileRoute {

const headers = Headers.from(null, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch bun.outOfMemory();

return bun.new(FileRoute, .{

.ref_count = .init(),

.server = opts.server,

.blob = blob,

.headers = headers,

.status_code = opts.status_code,

});

}

pub fn initFromBlob(blob: Blob, opts: InitOptions) *FileRoute {

const headers = Headers.from(null, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch bun.outOfMemory();

return bun.new(FileRoute, .{

.ref_count = .init(),

.server = opts.server,

.blob = blob,

.headers = headers,

.status_code = opts.status_code,

.has_last_modified_header = headers.get("last-modified") != null,

.has_content_length_header = headers.get("content-length") != null,

});

}

Spotted by Diamond

Is this helpful? React 👍 or 👎 to let us know.

src/bun.js/api/server/FileRoute.zig

Co-authored-by: graphite-app[bot] <96075541+graphite-app[bot]@users.noreply.github.com>

fix FileRoute response type

5010813

Jarred-Sumner added the codex label Jun 5, 2025 — with ChatGPT Connector

Jarred-Sumner and others added 9 commits June 5, 2025 03:06

bun scripts/glob-sources.mjs

4292a58

bun run zig-format

a7f8bca

Updated

a973467

bun scripts/glob-sources.mjs

26cbc43

bun run zig-format

32a6e87

bun run prettier

1943cda

More and better tests

74579aa

a

61e5219

Merge branch 'main' into codex/implement-fileroute-for-bun.serve

a6399e8

Jarred-Sumner marked this pull request as ready for review June 6, 2025 00:15

Merge branch 'main' into codex/implement-fileroute-for-bun.serve

dfe08a0

graphite-app bot reviewed Jun 6, 2025

View reviewed changes

src/bun.js/api/server/FileRoute.zig Show resolved Hide resolved

graphite-app bot reviewed Jun 6, 2025

View reviewed changes

src/bun.js/bindings/WTF.zig Outdated Show resolved Hide resolved

Jarred-Sumner added 3 commits June 5, 2025 18:37

Explicitly ref the FileReader. And add a micro optimization

2bf3129

Avoid negative numbers

a8ea971

The fix

8408c5b

graphite-app bot reviewed Jun 6, 2025

View reviewed changes

Jarred-Sumner added 5 commits June 5, 2025 21:03

Update FileRoute.zig

e2225ac

Update bun-serve-file.test.ts

85ef6bc

Update FileRoute.zig

dd20f58

Make it better at detecting when it ends

801ebd3

Update FileRoute.zig

0c5f847

graphite-app bot reviewed Jun 6, 2025

View reviewed changes

Try this

fa9bc58

Merge branch 'main' into codex/implement-fileroute-for-bun.serve

247dd8b

graphite-app bot reviewed Jun 7, 2025

View reviewed changes

dylan-conway and others added 4 commits June 6, 2025 21:31

Update FileRoute.zig

363a2a2

Merge branch 'main' into codex/implement-fileroute-for-bun.serve

587eb4b

Update src/bun.js/api/server/FileRoute.zig

8e34916

Co-authored-by: graphite-app[bot] <96075541+graphite-app[bot]@users.noreply.github.com>

bun run prettier

e136715

graphite-app bot reviewed Jun 7, 2025

View reviewed changes

Jarred-Sumner mentioned this pull request Jun 9, 2025

Introduce ahead of time bundling for HTML imports with bun build #20265

Merged

test

b35482b

graphite-app bot reviewed Jun 10, 2025

View reviewed changes

cirospaciari added 7 commits June 10, 2025 13:07

test

6aab118

test

fe8e084

more

0461f5e

opsie

adde716

opsie 2

862b0a0

fix ban words

fa1bc36

reduce file size so we dont timeout on CI

ccc9e90

graphite-app bot reviewed Jun 10, 2025

View reviewed changes

src/bun.js/api/server/FileRoute.zig Outdated Show resolved Hide resolved

dylan-conway and others added 2 commits June 10, 2025 16:30

Update src/bun.js/api/server/FileRoute.zig

ebe1375

Co-authored-by: graphite-app[bot] <96075541+graphite-app[bot]@users.noreply.github.com>

Update http.zig

9006146

dylan-conway approved these changes Jun 11, 2025

View reviewed changes

Merge branch 'main' into codex/implement-fileroute-for-bun.serve

e181163

dylan-conway merged commit 8750f0b into main Jun 11, 2025
5 of 7 checks passed

dylan-conway deleted the codex/implement-fileroute-for-bun.serve branch June 11, 2025 02:41

cirospaciari added 3 commits June 10, 2025 20:42

dont finish inside read wait for onReaderDonee

1c37bb5

fix windows

174f970

more fixes

b687868

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add FileRoute for serving files #20198

Add FileRoute for serving files #20198

Uh oh!

Jarred-Sumner commented Jun 5, 2025 •

edited by RiskyMH

Loading

Uh oh!

robobun commented Jun 5, 2025 •

edited

Loading

Uh oh!

RiskyMH commented Jun 5, 2025

Uh oh!

TiBianMod commented Jun 5, 2025

Uh oh!

Uh oh!

Uh oh!

graphite-app bot Jun 6, 2025

Uh oh!

graphite-app bot Jun 6, 2025

Uh oh!

Uh oh!

graphite-app bot Jun 6, 2025

Uh oh!

graphite-app bot Jun 7, 2025

Uh oh!

graphite-app bot Jun 7, 2025

Uh oh!

graphite-app bot Jun 10, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

	last_modified_buffer: [32]u8 = undefined,
	last_modified_buffer: [32]u8 = [_]u8{0} ** 32,

		response.body.value = .{ .Blob = blob.dupe() };
		const headers = Headers.from(response.init.headers, bun.default_allocator, .{ .body = &.{ .Blob = blob } }) catch bun.outOfMemory();

Add FileRoute for serving files #20198

Add FileRoute for serving files #20198

Uh oh!

Conversation

Jarred-Sumner commented Jun 5, 2025 • edited by RiskyMH Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Testing

Uh oh!

robobun commented Jun 5, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

RiskyMH commented Jun 5, 2025

Uh oh!

TiBianMod commented Jun 5, 2025

Uh oh!

Uh oh!

Uh oh!

graphite-app bot Jun 6, 2025

Choose a reason for hiding this comment

Uh oh!

graphite-app bot Jun 6, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

graphite-app bot Jun 6, 2025

Choose a reason for hiding this comment

Uh oh!

graphite-app bot Jun 7, 2025

Choose a reason for hiding this comment

Uh oh!

graphite-app bot Jun 7, 2025

Choose a reason for hiding this comment

Uh oh!

graphite-app bot Jun 10, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Jarred-Sumner commented Jun 5, 2025 •

edited by RiskyMH

Loading

robobun commented Jun 5, 2025 •

edited

Loading