fix(build): don't force -fuse-ld=lld on macOS#30871
Open
shguddn8591 wants to merge 7 commits into
Open
Conversation
## Summary Refactored `FFICallbackFunctionWrapper` to inherit from `WTF::ThreadSafeRefCounted` for safe cross-thread memory management. Guarded `JSFunction` execution in `FFI_Callback_threadsafe_call` by taking a protected `Ref` to prevent use-after-free crashes when JSC garbage collection runs concurrently with native thread callbacks. Fixed a debug-build assertion failure (`isASCII`) in `initializeInternalModuleFromDisk` caused by non-ASCII characters in `BUN_DYNAMIC_JS_LOAD_PATH` by using `WTF::String::fromUTF8` instead of `ASCIILiteral`. Added a high-stress test case simulating native thread callback execution with forced `Bun.gc(true)` to ensure thread safety. ## Test plan `bun bd test test/js/bun/ffi/cc.test.ts` passes reliably under high callback volume and forced garbage collection.
Homebrew's llvm@21 clang++ does not support -fuse-ld=lld on macOS. This caused 'invalid linker name' errors when running rust:check or debug builds. Only apply the LLD flag on non-Windows, non-Darwin platforms.
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
Contributor
WalkthroughRefactors FFI callback handling to a threadsafe refcounted wrapper capturing execution context, adds a skipped threadsafe JSCallback test exercising pthread callbacks with interleaved GC, omits the lld linker flag for Darwin in the Rust build, fixes an internal module path UTF-8 construction, and replaces AGENTS.md content with a reference to CLAUDE.md. ChangesFFI Callback Thread-Safety
Build Flags and Internal Module Loading
Developer Documentation
🚥 Pre-merge checks | ✅ 3 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Jarred-Sumner
requested changes
May 16, 2026
Jarred-Sumner
left a comment
Jarred-Sumner
left a comment
Collaborator
There was a problem hiding this comment.
The fix looks real but can you undo the AGENTS.md change? Most of the coding products follow symlinks.
Author
|
@Jarred-Sumner I have fulfilled your request. Sorry for the late correction. |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
scripts/build/rust.ts (1)
484-491: 🧹 Nitpick | 🔵 Trivial | ⚡ Quick winUpdate comment to document Darwin exclusion.
The comment at lines 484-488 explains why Windows is excluded from the
-fuse-ld=lldflag, but the code now also excludes Darwin (macOS). The comment should be updated to document both exclusions for future maintainability.📝 Proposed comment update
// Not on Windows: the per-target linker there is `link.exe` / `lld-link.exe` // (see `CARGO_TARGET_*_LINKER` below), which take `/X` args, not the GCC/clang // `-fuse-ld=`. RUSTFLAGS only reach *target* crates when `--target` is given, // and the `bun_bin` staticlib has no link step, so it's normally dead — but // if a target cdylib ever appears it'd fail with "could not open '-fuse-ld=lld'". + // Not on Darwin: macOS uses the system ld64 linker by default, and Homebrew's + // clang++ (llvm@21+) rejects the `-fuse-ld=lld` flag. if (!cfg.windows && !cfg.darwin) {🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@scripts/build/rust.ts` around lines 484 - 491, The comment explaining why we avoid adding `-fuse-ld=lld` only mentions Windows; update it to also document the Darwin exclusion by referencing the `cfg.windows` and `cfg.darwin` checks around the `rustflags.push(`-Clink-arg=-fuse-ld=lld`)` call so future readers understand both Windows and macOS are excluded (mentioning that macOS's linker/requirements differ and thus we skip adding `-fuse-ld=lld` on Darwin in the same way we skip it on Windows).
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@src/jsc/bindings/JSFFIFunction.cpp`:
- Around line 56-61: The constructor FFICallbackFunctionWrapper currently
dereferences globalObject->scriptExecutionContext() when initializing
m_contextIdentifier; add a defensive null check (or assert) for
globalObject->scriptExecutionContext() before using it so we don't crash when it
is nullptr—e.g., in the FFICallbackFunctionWrapper(JSC::JSFunction* function,
Zig::GlobalObject* globalObject) initializer or body, verify
scriptExecutionContext() is non-null and only call identifier() when valid (or
set m_contextIdentifier to a safe default/invalid value and log/assert) to match
the null-checking pattern used elsewhere (see uses in JSAbortSignal.cpp,
JSCallbackData.cpp, JSPerformanceObserver.cpp).
---
Outside diff comments:
In `@scripts/build/rust.ts`:
- Around line 484-491: The comment explaining why we avoid adding `-fuse-ld=lld`
only mentions Windows; update it to also document the Darwin exclusion by
referencing the `cfg.windows` and `cfg.darwin` checks around the
`rustflags.push(`-Clink-arg=-fuse-ld=lld`)` call so future readers understand
both Windows and macOS are excluded (mentioning that macOS's linker/requirements
differ and thus we skip adding `-fuse-ld=lld` on Darwin in the same way we skip
it on Windows).
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 8985a2a1-65fc-4bbd-ab8b-a1e6cec7835f
📒 Files selected for processing (6)
AGENTS.mdAGENTS.mdscripts/build/rust.tssrc/jsc/bindings/InternalModuleRegistry.cppsrc/jsc/bindings/JSFFIFunction.cpptest/js/bun/ffi/cc.test.ts
Comment on lines
56
to
61
| FFICallbackFunctionWrapper(JSC::JSFunction* function, Zig::GlobalObject* globalObject) | ||
| : m_function(globalObject->vm(), function) | ||
| , globalObject(globalObject->vm(), globalObject) | ||
| , m_contextIdentifier(globalObject->scriptExecutionContext()->identifier()) | ||
| { | ||
| } |
Contributor
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Check if scriptExecutionContext() can return nullptr in GlobalObject implementations
rg -nC3 'scriptExecutionContext\(\)' --type cpp | head -80Repository: oven-sh/bun
Length of output: 5768
🏁 Script executed:
# Find Zig::GlobalObject implementation and scriptExecutionContext() definition
fd -t f 'GlobalObject\|Zig' --type h --type cpp | head -30Repository: oven-sh/bun
Length of output: 224
🏁 Script executed:
# Search for Zig::GlobalObject class definition
rg -n 'class\s+GlobalObject' --type cpp --type h | head -20Repository: oven-sh/bun
Length of output: 1263
🏁 Script executed:
# Look for scriptExecutionContext() implementation in GlobalObject
rg -nB2 -A5 'GlobalObject.*scriptExecutionContext\(\)' --type cpp --type hRepository: oven-sh/bun
Length of output: 17952
🏁 Script executed:
# Check what calls FFICallbackFunctionWrapper constructor
rg -n 'FFICallbackFunctionWrapper' --type cpp | head -20Repository: oven-sh/bun
Length of output: 2756
Add null check for scriptExecutionContext() on line 59.
The constructor dereferences globalObject->scriptExecutionContext()->identifier() without null checking. The method returns a raw pointer that can be nullptr, as evidenced by the widespread null-checking pattern throughout the codebase (JSAbortSignal.cpp, JSCallbackData.cpp, JSPerformanceObserver.cpp, and others all defensively check this pointer). Add a null check or assert to prevent a potential crash during FFICallbackFunctionWrapper construction.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/jsc/bindings/JSFFIFunction.cpp` around lines 56 - 61, The constructor
FFICallbackFunctionWrapper currently dereferences
globalObject->scriptExecutionContext() when initializing m_contextIdentifier;
add a defensive null check (or assert) for
globalObject->scriptExecutionContext() before using it so we don't crash when it
is nullptr—e.g., in the FFICallbackFunctionWrapper(JSC::JSFunction* function,
Zig::GlobalObject* globalObject) initializer or body, verify
scriptExecutionContext() is non-null and only call identifier() when valid (or
set m_contextIdentifier to a safe default/invalid value and log/assert) to match
the null-checking pattern used elsewhere (see uses in JSAbortSignal.cpp,
JSCallbackData.cpp, JSPerformanceObserver.cpp).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Prevents the build system from forcing -fuse-ld=lld on macOS.
On macOS, Homebrew's llvm@21 clang++ does not support the -fuse-ld=lld flag, which caused the following error when running bun run rust:check or building debug binaries:
This change makes the LLD linker flag only apply to non-Windows, non-Darwin platforms.
How did you verify your code works?
Closes #30870