## Summary
Fixes#18413 - Empty chunked gzip responses were causing `Decompression
error: ShortRead`
## The Issue
When a server sends an empty response with `Content-Encoding: gzip` and
`Transfer-Encoding: chunked`, Bun was throwing a `ShortRead` error. This
occurred because the code was checking if `avail_in == 0` (no input
data) and immediately returning an error, without attempting to
decompress what could be a valid empty gzip stream.
## The Fix
Instead of checking `avail_in == 0` before calling `inflate()`, we now:
1. Always call `inflate()` even when `avail_in == 0`
2. Check the return code from `inflate()`
3. If it returns `BufError` with `avail_in == 0`, then we truly need
more data and return `ShortRead`
4. If it returns `StreamEnd`, it was a valid empty gzip stream and we
finish successfully
This approach correctly distinguishes between "no data yet" and "valid
empty gzip stream".
## Why This Works
- A valid empty gzip stream still has headers and trailers (~20 bytes)
- The zlib `inflate()` function can handle empty streams correctly
- `BufError` with `avail_in == 0` specifically means "need more input
data"
## Test Plan
✅ Added regression test in `test/regression/issue/18413.test.ts`
covering:
- Empty chunked gzip response
- Empty non-chunked gzip response
- Empty chunked response without gzip
✅ Verified all existing gzip-related tests still pass
✅ Tested with the original failing case from the issue
🤖 Generated with [Claude Code](https://claude.ai/code)
---------
Co-authored-by: Claude Bot <claude-bot@bun.sh>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
Co-authored-by: Ciro Spaciari <ciro.spaciari@gmail.com>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
* some work
* linux things
* linux things
* feat: tracestrings on Windows
* bwaa
* more work on the crash handler
* okay
* adgadsgbcxcv
* ya
* dsafds
* a
* wuh
* a
* bru h
* ok
* yay
* window
* alright
* oops
* yeah
* a
* a
* OOM handling
* fix on window
* Prepare for windows event loop
* More progress
* Update libuv.zig
* wip
* Make compiling each dependency a shell script
* Bump mimalloc
* Add the build scripts
* Update settings.json
* Fix a bunch of compiler warnings
* Remove more warnings
* more warnings
* cmake works
* Update JSSQLStatement.h
* put it in the zig file
* Fix usockets warnings
* Fixup
* Fix one of the compiler errors
* chunk
* draw the rest of the owl
* theres more
* Rename Process -> BunProcess
Works around a Windows issue
* Add musl polyfill for memmem on Windows
* More
* 12 mb
* Fix getenvZ
* fix variosu issues
* Add fast-ish path for bun install on Windows
* Update windows.zig
* Update windows.zig
* Fix build issue
* it works
* hmmm
* Rename file
* Fixups
* Update wtf-bindings.cpp
* Update src/bun.js/bindings/headers-handwritten.h
Co-authored-by: Dylan Conway <35280289+dylan-conway@users.noreply.github.com>
* further!
* more
* Update .gitignore
* hm
* quite a lot of fixes
* Update CMakeLists.txt
* zig fmt
* Many more things are starting to work.
* reb
* regenaret
* Update JSSink.h
* fixup
* fetch works
* Bun.serve() and much of the event loop works now
* Make require() work
* bun install progress
* more things work
* use less std.os
* Fixes
* small fixes
* Bump
* Bummp
* Undo that change
* We have to bump the version of Debian because libarchive has a higher minimum requirement
* ok
* some clenaup
* windows
* Update bun.zig
* fixup
* avoid duplicate symbols
* avoid undefined symbols
* bump
* Remove issue template for install
It's not used, and use the bug issue instead.
* Add types for cp and cpSync
* Add types for watchFile and unwatchFile
* Add bun-types to 'bun fmt' script
* Update nodejs compat docs cp/cpSync/watchFile/unwatchFile (#4525)
* feat(fetch) rejectUnauthorized and checkServerIdentity (#4514)
* enable root certs on fetch
* rebase
* fix lookup
* some fixes and improvements
* fmt
* more fixes
* more fixes
* check detached onHandshake
* fix promise case
* fix cert non-Native
* add fetch tls tests
* more one test
* churn
* Update feature_flags.zig
* Update response.zig
* Revert "avoid undefined symbols"
This reverts commit ca835b726f.
* Revert "avoid duplicate symbols"
This reverts commit 4ac6ca8700.
* Update feature_flags.zig
* Set permissions
* more
* Update mimalloc
* Fix sqlite test failures
* Fix some test failures
* Make sure we remove libusockets is removed
* hm
* [dave]: fix webcrypto crash
* bump
* Update index.ts
* windows zig compiles
* cmake on mac works
* progress
* yay
* bun run build
* fix
* ok
* oops
* asdfasfdafdsafda
* fghjkl
* git ignore
* wow
* Process -> BunProcess
* hmm
* blah
* finalize merge
* now it only has linker errors on mac
* sdfadsf
* g
* getting farther
* sxdcvbnmk,
* adfhjskfjdhkas
* a
* fgh
* update build dot zig
* asdfg
* theoretical -DCANARY flag we can use
* asdf
* cool
* okay
* colorterm
* New build workflow
* Fix script
* Use sudo
* More sudo
* Tweak dependencies
* Another sudo attempt
* Tweak script
* 16.0 -> 16
* Tweak script
* Tweak script
* Tweak script
* Tweak script
* Tweak script
* bun install
* ssh into github actions
* add more to ssh
* Fix postinstal
* Skip llvm
* New dockerfile
* Build
* More changes to Dockerfile
* chaos chaos chaos
* okay
* a
* more cmake nonsense
* add unified sources code (does not work)
* stuff
* prepare for CI builds
* ok
* yay
* yeah
* make this more stable simply by trying again if it fails, 5 times, then lose. it fixes the stability issue i was running into L O L
* messing with ci
* x
* a
* clean dependencies before build
* oops
* this is not going to work but its closer
* not gonna work either
* a
* a
* did i do it
* a
* a
* work around weird fs+Bun.build issues
* properly pass debug flag correctly
* idk im sorry
* lose
* maybe
* run the tests please
* a
* fix zlib script
* a
* hi
* prevent stupid ci issue
* i totally didnt leave in a syntax error on cmakelists
* a
* lol
* relax
* 😭
* a
* SO SILLY
* 😡 one line mistake
* one character diff
* fix linking symbols missing
* work on dependency scripts
* does this work now?
* fix mac build
* a
* bump!
* woops
* add macos baseline build
* .
* fix sqlite and also enable $assert/$debug support in builtin functions
* okay
* oops
* zig upgrade lol
* Merge
* Fix spawn test issue
* Set a timeout
* yeah
* etc
* mi
---------
Co-authored-by: Jarred Sumner <709451+Jarred-Sumner@users.noreply.github.com>
Co-authored-by: Dylan Conway <dylan.conway567@gmail.com>
Co-authored-by: Dylan Conway <35280289+dylan-conway@users.noreply.github.com>
Co-authored-by: Ashcon Partovi <ashcon@partovi.net>
Co-authored-by: Birk Skyum <74932975+birkskyum@users.noreply.github.com>
Co-authored-by: dave caruso <me@paperdave.net>
* make pending_response and metdata life cycle more clear
* typo
* WIP: memory investigation
* check zlib and fix zlib
* use state allocator for metadata
* remove postBodyProcess
* undo some test things
* fix race condition
* fix removing compressed header
* some extra checks
* remove arenas on zlib and comment repoter.assert because of toOwnedSliceZ
---------
Co-authored-by: Jarred Sumner <jarred@jarredsumner.com>
* streams non compressed data in 64kb chunks (at least)
* fmt
* wip remove pause
* fix default streaming and buffering
* fix atomic lags
* fix size
* make chunked encoding work again (WIP streaming chunked)
* WIP: chunked encoding streaming
* fix end of streamings
* working streaming + compression
* add fixes + tests
* fmt + fix proxy
* fix oopsies
* codegen after merge
* fmt + fixes
* more fixes
* more fixes and logs
* avoid double free
* check empty before pop
* check empty on pop
* fix copy to real when complete
* remove unnecessary logs
* better has_schedule_callback swap, body locked size helper, remove isEmpty from unbounded_queue pop
* fix response ref, fix body_size
* add deflate support, fix error throw, add more tests
* codegen after merge
* remove logs, add connection close test
* fix macOS build
* fix redirect error option
* make body_size more clear
* support new Reponse(response)
* toString DOMWrapper objects properly instead of supporting response in Response constructor
* ignore headers with no name, add more tests
* oops
* handle transform with fetch
* add gz image stream test
* remove duplicate test
* fix missing chunk on macOS under pressure
* oops include all OS
* some fixes
* compare buffers instead of sizes
* refactor err.err and protect it
- This removes memory pooling from the HTTP client which sometimes caused invalid memory to be written to the response body.
- This adds a fast path for small HTTP/HTTPS responses that makes it a single memory allocation for the response body, instead of copying & allocating a temporary buffer
cc @Electroid
