[trivikr]

Fixes: #66841

registerCallback() returns a plain bigint address, which doesn't keep the owning DynamicLibrary reachable. If user code dropped the library without calling close(), GC ran ~DynamicLibrary() and freed every ffi_closure. Native code still holding a callback pointer then segfaulted, even though no documented rule was broken.

This PR makes each FFICallback hold its owning library through a BaseObjectPtr, so the library stays strong until its last callback is unregistered or close() is called. Callbacks after unrefCallback() still count, since their trampolines remain live native memory. Because callbacks and library now reference each other, an Environment cleanup hook releases the callbacks so the library is still freed at teardown (worker or process exit).

---

Assisted-by: claude:opus-5.5

[nodejs-github-bot]

Review requested:

• @nodejs/ffi

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.50%. Comparing base (1a885ca) to head (7e8c874).
⚠️ Report is 8 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #66842      +/-   ##
==========================================
- Coverage   90.51%   90.50%   -0.01%     
==========================================
  Files         797      797              
  Lines      278069   278078       +9     
  Branches    53416    53418       +2     
==========================================
- Hits       251690   251671      -19     
- Misses      16760    16800      +40     
+ Partials     9619     9607      -12     

Files with missing lines	Coverage Δ
src/node_ffi.cc	72.90% <100.00%> (+0.27%)	⬆️
src/node_ffi.h	77.77% <ø> (ø)

... and 34 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[addaleax]

Just going to point out that there is an existing for this problem in the form of using BaseObjectPtr<DynamicLibrary> owner; instead of DynamicLibrary* owner; in struct FFICallback

[trivikr]

Just going to point out that there is an existing for this problem in the form of using BaseObjectPtr<DynamicLibrary> owner; instead of DynamicLibrary* owner; in struct FFICallback

Thanks. Switched to BaseObjectPtr. The resulting cycle aborted in ~Realm on worker exit, so the library now releases its callbacks from a cleanup hook. Added a test for that case.