chip-tool segfault on Darwin

[mrjerryjohns]

Problem

When I run scripts/tests/test_suite.sh on Darwin, after about 5-10 iterations, the test crashes with a segmentation fault.

Crash backtrace:

Thread 1 Crashed:: Dispatch queue: com.zigbee.chip.framework.controller.workqueue
0   chip-tool                     	0x0000000103b2abc8 chip::TransportMgrBase::SendMessage(chip::Transport::PeerAddress const&, chip::System::PacketBufferHandle&&) + 24 (TransportMgrBase.cpp:27)
1   chip-tool                     	0x0000000103b273b5 chip::SecureSessionMgr::SendMessage(chip::SecureSessionHandle, chip::PayloadHeader&, chip::PacketHeader&, chip::System::PacketBufferHandle&&, chip::EncryptedPacketBufferHandle*, chip::SecureSessionMgr::EncryptionState) + 1125 (SecureSessionMgr.cpp:208)
2   chip-tool                     	0x0000000103b26f45 chip::SecureSessionMgr::SendMessage(chip::SecureSessionHandle, chip::PayloadHeader&, chip::System::PacketBufferHandle&&, chip::EncryptedPacketBufferHandle*) + 149 (SecureSessionMgr.cpp:129)
3   chip-tool                     	0x0000000103b09acc chip::Messaging::ApplicationExchangeDispatch::SendMessageImpl(chip::SecureSessionHandle, chip::PayloadHeader&, chip::System::PacketBufferHandle&&, chip::EncryptedPacketBufferHandle*) + 108 (ApplicationExchangeDispatch.cpp:33)
4   chip-tool                     	0x0000000103b0b3f8 chip::Messaging::ExchangeMessageDispatch::SendMessage(chip::SecureSessionHandle, unsigned short, bool, chip::Messaging::ReliableMessageContext*, bool, chip::Protocols::Id, unsigned char, chip::System::PacketBufferHandle&&) + 824 (ExchangeMessageDispatch.cpp:97)
5   chip-tool                     	0x0000000103b0a127 chip::Messaging::ExchangeContext::SendMessage(chip::Protocols::Id, unsigned char, chip::System::PacketBufferHandle&&, chip::BitFlags<chip::Messaging::SendMessageFlags, unsigned short> const&) + 679 (ExchangeContext.cpp:124)
6   chip-tool                     	0x0000000103b41a68 int chip::Messaging::ExchangeContext::SendMessage<chip::Protocols::SecureChannel::MsgType, void>(chip::Protocols::SecureChannel::MsgType, chip::System::PacketBufferHandle&&, chip::BitFlags<chip::Messaging::SendMessageFlags, unsigned short> const&) + 88 (ExchangeContext.h:108)

Thread 0:
23  chip-tool                     	0x0000000103a14483 Commands::~Commands() + 83 (Commands.h:26)
24  chip-tool                     	0x0000000103a13475 Commands::~Commands() + 21 (Commands.h:26)
25  chip-tool                     	0x0000000103a133ff main + 191 (main.cpp:44)
26  libdyld.dylib                 	0x00007fff20380621 start + 1

This looks like another race condition where the main thread had already torn down the stack, but a packet came in on the chip dispatch work queue and tried to access TransportMgr, which had by now, been destructed.

[woody-apple]

Not showing up in CI, or in automation (internal), closing.