Saw some failure in the
PtrQueueBufferAllocatorTest stress_free_list_allocator test.

# Internal Error (/home/runner/work/jdk/jdk/jdk/test/hotspot/gtest/gc/shared/test_ptrQueueBufferAllocator.cpp:115), pid=1882, tid=1882
# assert(_completed_list.empty()) failed: completed list not empty
#
# JRE version: OpenJDK Runtime Environment (18.0) (fastdebug build 18-internal+0-tschatzl-b8e67fe8065556ad603220ccec731de25ef55b48)
# Java VM: OpenJDK Server VM (fastdebug 18-internal+0-tschatzl-b8e67fe8065556ad603220ccec731de25ef55b48, mixed mode, tiered, g1 gc, linux-x86)
# Problematic frame:
# V [libjvm.so+0x501237] PtrQueueBufferAllocatorTest_stress_free_list_allocator_vm_Test::TestBody()+0xfe7

This seems to be a test issue: there is no provision that guarantees that the "ProcessorThread" pops one element after getting the shutdown command via a flag.

I.e. this code:

  virtual void main_run() {
    while (true) {
      BufferNode* node = _cbl->pop();
      if (node != NULL) {
        _allocator->release(node);
      } else if (!Atomic::load_acquire(_continue_running)) {
        return;
      }
      ThreadBlockInVM tbiv(this); // Safepoint check.
    }
  }

exits prematurely if the _continue_running flag is set but this particular loop iteration did not manage to pop a element.

This is a failure from github actions, so no particular other artifacts available than the attached hs_err file.

Seen exactly once on linux-x86, but this does not seem platform specific.