## Summary
This PR fixes multiple HTTP/2 protocol compliance issues that were
causing stream errors with various HTTP/2 clients (Fauna, gRPC/Connect,
etc.).
fixes https://github.com/oven-sh/bun/issues/12544
fixes https://github.com/oven-sh/bun/issues/25589
### Key Fixes
**Window Size and Settings Handling**
- Fix initial stream window size to use `DEFAULT_WINDOW_SIZE` until
`SETTINGS_ACK` is received
- Per RFC 7540 Section 6.5.1: The sender can only rely on settings being
applied AFTER receiving `SETTINGS_ACK`
- Properly adjust existing stream windows when `INITIAL_WINDOW_SIZE`
setting changes (RFC 7540 Section 6.9.2)
**Header List Size Enforcement**
- Implement `maxHeaderListSize` checking per RFC 7540 Section 6.5.2
- Track cumulative header list size using HPACK entry overhead (32 bytes
per RFC 7541 Section 4.1)
- Reject streams with `ENHANCE_YOUR_CALM` when header list exceeds
configured limit
**Custom Settings Support**
- Add validation for `customSettings` option (up to 10 custom settings,
matching Node.js `MAX_ADDITIONAL_SETTINGS`)
- Validate setting IDs are in range `[0, 0xFFFF]` per RFC 7540
- Validate setting values are in range `[0, 2^32-1]`
**Settings Validation Improvements**
- Use float comparison for settings validation to handle large values
correctly (was using `toInt32()` which truncates)
- Use proper `HTTP2_INVALID_SETTING_VALUE_RangeError` error codes for
Node.js compatibility
**BufferFallbackAllocator** - New allocator that tries a provided buffer
first, falls back to heap:
- Similar to `std.heap.stackFallback` but accepts external buffer slice
- Used with `shared_request_buffer` (16KB threadlocal) for common cases
- Falls back to `bun.default_allocator` for large headers
## Test Plan
- [x] `bun bd` compiles successfully
- [x] Node.js HTTP/2 tests pass: `bun bd
test/js/node/test/parallel/test-http2-connect.js`
- [x] New regression tests for frame size issues: `bun bd test
test/regression/issue/25589.test.ts`
- [x] HTTP/2 continuation tests: `bun bd test
test/js/node/http2/node-http2-continuation.test.ts`
---------
Co-authored-by: Claude Opus 4.5 <noreply@anthropic.com>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: Claude Bot <claude-bot@bun.sh>
Co-authored-by: Jarred Sumner <jarred@jarredsumner.com>
## Summary
Removes the `MemoryReportingAllocator` wrapper and simplifies
`fetch.zig` to use `bun.default_allocator` directly. The
`MemoryReportingAllocator` was wrapping `bun.default_allocator` to track
memory usage for reporting to the VM, but this added unnecessary
complexity and indirection without meaningful benefit.
## Changes
**Deleted:**
- `src/allocators/MemoryReportingAllocator.zig` (96 lines)
**Modified `src/bun.js/webcore/fetch.zig`:**
- Removed `memory_reporter: *bun.MemoryReportingAllocator` field from
`FetchTasklet` struct
- Removed `memory_reporter: *bun.MemoryReportingAllocator` field from
`FetchOptions` struct
- Replaced all `this.memory_reporter.allocator()` calls with
`bun.default_allocator`
- Removed all `this.memory_reporter.discard()` calls (no longer needed)
- Simplified `fetch()` function by removing memory reporter
allocation/wrapping/cleanup code
- Updated `deinit()` and `clearData()` to use `bun.default_allocator`
directly
**Cleanup:**
- Removed `MemoryReportingAllocator` export from `src/allocators.zig`
- Removed `MemoryReportingAllocator` export from `src/bun.zig`
- Removed `bun.MemoryReportingAllocator.isInstance()` check from
`src/safety/alloc.zig`
## Testing
- ✅ Builds successfully with `bun bd`
- All fetch operations now use `bun.default_allocator` directly
## Impact
- **Net -116 lines** of code
- Eliminates allocator wrapper overhead in fetch operations
- Simplifies memory management code
- No functional changes to fetch behavior
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Co-authored-by: Claude Bot <claude-bot@bun.sh>
Co-authored-by: Claude <noreply@anthropic.com>
tests not in `test/no-validate-leaksan.txt` will run with leaksanitizer
in CI
leaks documented in `test/leaksan.supp` will not cause a test failure
-- notes about leaksanitizer
- will not catch garbage collected objects accumulated during
long-running processes
- will not catch js objects (eg a strong held to a promise)
- will catch native calls to `malloc` not `free`d
- will catch allocations made in C, Zig, C++, libc, dependencies,
dlopen'd
---------
Co-authored-by: Jarred Sumner <jarred@jarredsumner.com>
* Define a generic allocator interface to enable static polymorphism for
allocators (see `GenericAllocator` in `src/allocators.zig`). Note that
`std.mem.Allocator` itself is considered a generic allocator.
* Add utilities to `bun.allocators` for working with generic allocators.
* Add a new namespace, `bun.memory`, with basic utilities for working
with memory and objects (`create`, `destroy`, `initDefault`, `deinit`).
* Add `bun.DefaultAllocator`, a zero-sized generic allocator type whose
`allocator` method simply returns `bun.default_allocator`.
* Implement the generic allocator interface in `AllocationScope` and
`MimallocArena`.
* Improve `bun.threading.GuardedValue` (now `bun.threading.Guarded`).
* Improve `bun.safety.AllocPtr` (now `bun.safety.CheckedAllocator`).
(For internal tracking: fixes STAB-1085, STAB-1086, STAB-1087,
STAB-1088, STAB-1089, STAB-1090, STAB-1091)
Replace `catch bun.outOfMemory()`, which can accidentally catch
non-OOM-related errors, with either `bun.handleOom` or a manual `catch
|err| switch (err)`.
(For internal tracking: fixes STAB-1070)
---------
Co-authored-by: Dylan Conway <dylan.conway567@gmail.com>
We can't use `std.heap.c_allocator` as `z_allocator`; it doesn't
zero-initialize the memory. This PR adds a fallback implementation.
This PR also makes Bun compile successfully with `use_mimalloc` set to
false. More work is likely necessary to make it function correctly in
this case, but it should at least compile.
(For internal tracking: fixes STAB-978, STAB-979)
* Revert "fix!: do not lookup cwd in which (#9691)"
This reverts commit 4869ebff24.
* fix which implementation to be more accurate
* t
* which tests windows
* [autofix.ci] apply automated fixes
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
* WIP sync close (shows ref count bug in stream)
* fix closing on PipeWriter and PipeReader
* remove old todos
* join
* Some shell changes
at least it compiles
* fix some compile errors
* fix ref/unref server on windows
* actually use the ref count in this places
* make windows compile again
* more tests passing
* Make shell compile again
* Slowly remove some `@panic("TODO SHELL")`
* Eliminate `@panic("TODO SHELL")` for BufferedWriter
* Holy cleansing of `@panic("TODO SHELL")`
at least it compiles now
* Okay now the shell compiles, but segfaults
* Fix compiler errors
* more stable stream and now Content-Range pass
* make windows compile again
* revert stuff until the fix is actually ready
* revert onDone thing
* Fix buffered writer for shell
* Fix buffered writer + shell/subproc.zig and windows build
* Fix for #8982 got lost in the merge
* Actually buffer subproc output
* Fix some stuff shell
* oops
* fix context deinit
* fix renderMissing
* shell: Fix array buffer
* more stable streams (#9053)
fix stream ref counting
* wip
* Remove `@panic("TODO")` on shell event loop tasks and Redirect open flags got lost in merge
* Support redirects
* fixes
cc @cirospaciari
* Update ReadableStreamInternals.ts
* Fix spurious error
* Update stream.js
* leak
* Fix UAF
cc @cirospaciari
* Fix memory leaks
* HOLY FUCK big refactor
* misc cleanup
* shell: Fix a bunch of tests
* clean up
* gitignore: fix ending newline
* get windows compiling again
* tidy
* hide linker warn with icu
* closeIfPossible
* Better leak test
* Fix forgetting to decrement reference count
* Update stdio.zig
* Fix shell windows build
* Stupid unreachable
* Woops
* basic echo hi works on windows
* Fix flaky test on Windows
* Fix windows regression in Bun.main (#9156)
* Fix windows regression in Bun.main
* Handle invalid handles
* Fix flaky test
* Better launch config
* Fixup
* Make this test less flaky on Windows
* Fixup
* Cygwin
* Support signal codes in subprocess.kill(), resolve file path
* Treat null as ignore
* Ignore carriage returns
* Fixup
* shell: Fix IOWriter bug
* shell: Use custom `open()`/`openat()`
* windows shell subproc works
* zack commit
* I think I understand WindowsStreamingWriter
* fix thing
* why were we doing this in tests
* shell: Fix rm
* shell: Add rm -rf node_modules/ test
* shell: use `.runAsTest()` in some places to make it easier to determine which test failed
* [autofix.ci] apply automated fixes
* woopsie
* Various changes
* Fix
* shell: abstract output task logic
* shell: mkdir builtin
* fixup
* stuff
* shell: Make writing length of 0 in IOWriter immediately resolve
* shell: Implement `touch`
* shell: basic `cat` working
* Make it compile on windows
* shell: Fix IOReader bug
* [autofix.ci] apply automated fixes
* fix windows kill on subprocess/process
* fix dns tests to match behavior on windows (same as nodejs)
* fix windows ci
* again
* move `close_handle` to flags in `PipeWriter` and fix shell hanging
* Fix `ls` not giving non-zero exit code on error
* Handle edgecase in is_atty
* Fix writer.flush() when there's no data
* Fix some tests
* Disable uv_unref on uv_process_t on Windows, for now.
* fix writer.end
* fix stdout.write
* fix child-process on win32
* Make this test less flaky on Windows
* Add assertion
* Make these the same
* Make it pass on windows
* Don't commit
* Log the test name
* Make this test less flaky on windows
* Make this test less flaky on windows
* Print which test is taking awhile in the runner
* fixups
* Fixups
* Add some assertions
* Bring back test concurrency
* shell: bring back redirect stdin
* make it compile again cc @zackradisic
* initialize env map with capacity
* some fixes
* cleanup
* oops
* fix leak, fix done
* fix unconsumedPromises on events
* always run expect
* Update child_process.test.ts
* fix reading special files
* Fix a test
* Deflake this test
* Make these comparisons easier
* Won't really fix it but slightly cleaner
* Update serve.test.ts
* Make the checks for if the body is already used more resilient
* Move this to the harness
* Make this test not hang in development
* Fix this test
* Make the logs better
* zero init some things
* Make this test better
* Fix readSocket
* Parallelize this test
* Handle EPipe and avoid big data
* This was a mistake
* Fix a bunch of things
* Fix memory leak
* Avoid sigpipe + optimize + delete dead code
* Make this take less time
* Make it bigger
* Remove some redundant code
* Update process.zig
* Merge and hopefully don't breka things along teh way
* Silence build warning
* Uncomment on posix
* Skip test on windows
* windows
* Cleanup test
* Update
* Deflake
* always
* less flaky test
* [autofix.ci] apply automated fixes
* logs
* fix uaf on shell IOReader
* stuff to make it work with mini event loop
* fix 2 double free scenarios, support redirections on windows
* shell: Make `1>&2` and `2>&1` work with libuv
* yoops
* Partial fix
* Partial fix
* fix build
* fix build
* ok
* Make a couple shell tests pass
* More logging
* fix
* fix
* Fix build issue
* more tests pass
* Deflake
* Deflake
* Use Output.panic instead of garbled text
* Formatting
* Introduce `bun.sys.File`, use it for `Output.Source.StreamType`, fix nested Output.scoped() calls, use Win32 `ReadFile` API for reading when it's not a libuv file descriptor.
This lets us avoid the subtle usages of `unreachable` in std.os when writing to stdout/stderr.
Previously, we were initializing the libuv loop immediately at launch due to checking for the existence of a bun build --compile'd executable. When the file descriptor is not from libuv, it's just overhead to use libuv
cc @paperdave, please tell me if Iany of that is incorrect or if you think this is a bad idea.
* Fix closing undefined memory file descriptors in spawn
cc @zackradisic
* pause instead of close
* Fix poorly-written test
* We don't need big numbers for this test
* sad workaround
* fixup
* Clearer error handling for this test
* Fix incorrect test
@electroid when ReadableStream isn't closed, hanging is the correct behavior when consuming buffered data. We cannot know if the buffered data is finished if the stream never closes.
* Fix build
* Remove known failing on windows
* Deflake
* Mark no longer failing
* show all the failing tests
* Sort the list of tests
* fix argument handling
* dont show "posix_spawn" as an error code on windows
* make bun-upgrade.test.ts pass on windows
* fix bunx and bun create again sorry
* a
* fix invalidexe because we should not be running javascript files as if they were exes
* Concurrency in test runner + better logging
* Revert "fix invalidexe because we should not be running javascript files as if they were exes"
This reverts commit da47cf8247.
* WIP: Unix fixes (#9322)
* wip
* [autofix.ci] apply automated fixes
* wip 2
* [autofix.ci] apply automated fixes
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: Jarred Sumner <jarred@jarredsumner.com>
* Update runner.node.mjs
* Update runner.node.mjs
* Document some environment variables
* shell: Make `Response` work with builtins
* Make it compile
* make pwd test pass
* [autofix.ci] apply automated fixes
* Fix printing garbage for source code previews
* Update javascript.zig
* Fix posix test failures
* Fix signal dispatch
cc @paperdave. Signals can be run from any thread. This causes an assertion failure when the receiving thread happens to not be the main thread. Easiest to reproduce on linux when you spawn 100 short-lived processes at once.
* windows
---------
Co-authored-by: cirospaciari <ciro.spaciari@gmail.com>
Co-authored-by: Zack Radisic <56137411+zackradisic@users.noreply.github.com>
Co-authored-by: Zack Radisic <zackradisic@Zacks-MBP-2.attlocal.net>
Co-authored-by: Jarred Sumner <709451+Jarred-Sumner@users.noreply.github.com>
Co-authored-by: Meghan Denny <meghan@bun.sh>
Co-authored-by: Zack Radisic <zack@theradisic.com>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: dave caruso <me@paperdave.net>
Co-authored-by: Dylan Conway <dylan.conway567@gmail.com>
* random acts of cleanup to tests
* nerf the watchFile test, is CI slow?
* more things
* [autofix.ci] apply automated fixes
* a
* remove a log
---------
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
* work on bundler
* a
* YAAAAYYAYAYAYYAYA
* get some more bundler tests working
* Update src/bundler/bundle_v2.zig
* rev
* ok
* i converted the cmakelists into LF
* personal review
* we didnt win
* okey they pass
* revert :(
* a
---------
Co-authored-by: Jarred Sumner <jarred@jarredsumner.com>
* chore: upgrade zig to 0.12.0-dev.1828+225fe6ddb
* open as iterable
* fix building identifier cache
* fix windows build
* fix linux build
* fix linux build
* very very wip
* almost ready to fix the errors
* Update identity_context.zig
* Update base.zig
* [bun test] It runs successfully
* Remove unnecessary call
* [Bun.js] Improve JS <> Zig unicode string interop
This fixes longstanding unicode bugs with `console.log` & `fetch`.
I believe @evanwashere reported this first awhile ago
* [Bun.js] Implement `Object.is()` binding and a way to set a timeout for script execution
* Update PLCrashReport.zig
* [Bun.js] Make `console.log` more closely match Node.js and Deno
* [Bun.js] Implement formatting specifier for console.*
* Implement `console.clear()`
* bug fix
* Support console.clear()
* Buffer stderr
* [bun test] Begin implementing Node.js `fs`
* Update darwin_c.zig
* Implement more of `fs`
* `mkdir`, `mkdir` recursive, `mkdtemp`
* `open`, `read` (and pread)
* Move some things into more files
* Implement readdir
* `readFile`, `readLink`, and `realpath`
* `writeFile`, `symlink`, `chown`, `rename`, `stat`, `unlink`, `truncate`
* `lutimes`
* Implement `SystemError` and begin wiring up the `fs` module
* `"fs"` - Most of the arguments / validation
* `fs` - Rest of the arguments / validations
* Begin wiring up the `fs` module
* Fix all the build errors
* support printing typed arrays in console.log
* It...works?
* Support `require("fs")`, `import fs from 'fs';`, `import * as fs from 'fs'`
* Fix a couple bugs
* get rid of the crash reporter for now
* Update fs.exports.js
* [bun.js] slight improvement to startup time
* [bun.js] Improve error message printing
* [Bun.js] Add `Bun.gc()` to run the garbage collector manually and report heap size
* [Bun.js] Add Bun.generateHeapSnapshot to return what JS types are using memory
* [Bun.js] Add `Bun.shrink()` to tell JSC to shrink the VM size
* Improve encoding reader
* [bun.js] Improve callback & microtask performance
* Update node_fs.zig
* Implement `console.assert`
* simple test
* [Bun.js] Prepare for multiple globals/realms to support testing
* Create callbacks-overhead.mjs
* Update http.zig
* [Bun.js] Implement `queueMicrotask`
* Add test for queueMicrotask
* 😪
* [Bun.js] Implement `process.versions`, `process.pid`, `process.ppid`, `process.nextTick`, `process.versions`,
* Implement `process.env.toJSON()`
* [Bun.js] Improve performance of `fs.existsSync`
* 💅
* [Bun.js] Implement `process.chdir(str)` and `process.cwd()`, support up to 4 args in `process.nextTick`
* Make creating Zig::Process lazy
* Split processi nto separte file
* [Bun.js] Node.js Streams - Part 1/?
* [Bun.js] Node.js streams 2/?
* WIP streams
* fix crash
* Reduce allocations in many places
* swap
* Make `bun` start 2ms faster
* Always use an apiLock()
* libBacktrace doesn't really work yet
* Fix crash in the upgrade checker
* Clean up code for importing the runtime when not bundling
* 📷
* Update linker.zig
* 68!
* backtrace
* no, really backtrace
* Fix
* Linux fixes
* Fixes on Linux
* Update mimalloc
* [bun test] Automatically scan for {.test,_test,.spec,_spec}.{jsx,tsx,js,cts,mts,ts,cjs}
* Remove some dead code
* 💀 code
* Fix the zig tests
* [JS Printer] Print integers faster & less scientific notation on decimals
* 💀 dead code
* skip
* Run all the unit tests
