Crashed: GC_is_marked for IOS #154

Activity

1. ericmsauer commented on Jan 21, 2026

   

   ericmsauer

   on Jan 21, 2026

   More actions

   We hit this issue as well. Were you able to resolve your issue?

   YAHH: 1.11.4
   Unity: 6.0.58f2

   We believe it might have to do with some il2cpp stripping, did you add these libraries to your link.xml?
2. PawelMilaszewicz commented on Sep 1, 2026

   

   PawelMilaszewicz

   on Sep 1, 2026

   Last edited by PawelMilaszewicz

   More actions

   We're hitting this same crash in production on YetAnotherHttpHandler 1.11.4, Unity 6000.3.17f1, iOS 26.x. It's the top iOS fatal in our app (that does gRPC over HTTP/2).

   Our variant faults one frame earlier than the GC allocation in the original report, but it's the same thread-attach on the tokio worker:

   Crashed: tokio-runtime-worker
   0  pthread_getspecific
   1  il2cpp::vm::ScopedThreadAttacher::ScopedThreadAttacher()  (ScopedThreadAttacher.cpp:10)
   2  ReversePInvokeWrapper.NativeHttpHandlerCore_OnReceive     (Cysharp.Net.Http.YetAnotherHttpHandler.cpp)
   3  yaha_native::binding::yaha_request_begin::{{closure}}
      … tokio multi_thread worker → blocking::pool::Inner::run …
   

   Both stacks are NativeHttpHandlerCore_OnReceive → ScopedThreadAttacher; #154 faults inside the GC allocation that Thread::Attach performs, ours in the pthread_getspecific right before it. Same root: a tokio worker invokes the OnReceive reverse-P/Invoke and il2cpp attaches the native thread while the app is backgrounding.

   Trigger, from breadcrumbs (100% processState: BACKGROUND, consistent with #142):

   • The crash fires ~1 second after OnApplicationPause(true) (app sent to background), while there is in-flight HTTP/2 activity — often right after a reconnect, when YAHA is completing failed requests.

   Things we tried that did not fix it (corroborating @utkayd in #142 that pausing/resuming is ineffective):

   • On background we Dispose() the YetAnotherHttpHandler, which drops the last ref so NativeRuntime is released and yaha_dispose_runtime runs. This removed a batch of related ObjectDisposedExceptions but did not eliminate this fatal.
   • We think that's because yaha_dispose_runtime is a plain, unbounded Runtime drop performed on the main thread during OnApplicationPause: it can't interrupt a worker already inside the OnReceive callback, and iOS can suspend the process before the drop finishes joining the workers.

   The core issue seems to be that the OnReceive/OnComplete reverse-P/Invoke callbacks attach a thread and allocate through the GC, which isn't safe while iOS is suspending the process. Two things that would help from the library side:

   1. A bounded runtime shutdown (e.g. Runtime::shutdown_timeout / shutdown_background) exposed via FFI, so an app can hard-stop the tokio runtime on background without an unbounded main-thread join.
   2. Guarding the reverse-P/Invoke callbacks to no-op when the runtime/handler is being torn down, rather than attaching a thread.

   Happy to share more crash instances or test a build. Thanks for the library!
3. alvinsay commented on Sep 15, 2026

   

   alvinsay

   on Sep 15, 2026

   More actions

   We hit the same issue.

   Just made a PR. #171