Session restore causes crash on Linux

[mkieweg]

There seems to be an issue in the Nakama cpp wrapper library for Linux. Restoring a session using the Nakama wrapper results in the following uncaught exception:

[2021.12.14-19.24.25:569][968]LogTemp: [Nakama::RestClient::RestClient] Created. NakamaSdkVersion: 2.5.0
[2021.12.14-19.24.25:569][968]ProjectVNakama: existing session found, restoring session
[2021.12.14-19.24.25:570][968]ProjectVNakama: Stored session token in GameInstance
[2021.12.14-19.24.25:570][968]ProjectVNakama: Handling nakama session, creating Real Time Client instance, assigning listeners and triggering connect (Owner: BP_GameModeProcedural_C_2147482195)
[2021.12.14-19.24.25:570][968]LogTemp: [NRtClient::NRtClient] Created
[2021.12.14-19.24.25:570][968]LogTemp: [NRtClient::connect] ...
CommonUnixCrashHandler: Signal=5
[2021.12.14-19.24.25:602][968]LogCore: === Critical error: ===
Unhandled Exception: SIGTRAP: trace trap

[2021.12.14-19.24.25:602][968]LogCore: Fatal error!

0x00007f7082cc624b libpthread.so.0!raise(+0xcb)
0x00007f7081f70e52 libnakama-cpp.so!UnknownFunction(0x123e51)
0x00007f7081f53ae8 libnakama-cpp.so!_ZNSt16_Sp_counted_baseILN9__gnu_cxx12_Lock_policyE2EE10_M_releaseEv(+0x47)
0x00007f7082141c3c libnakama-cpp.so!UnknownFunction(0x2f4c3b)
0x00007f708223e396 libnakama-cpp.so!UnknownFunction(0x3f1395)
0x00007f70821aa593 libnakama-cpp.so!_ZN5boost4asio6detail9scheduler3runERNS_6system10error_codeE(+0x4a2)
0x00007f708223f116 libnakama-cpp.so!UnknownFunction(0x3f2115)
0x00007f7082193404 libnakama-cpp.so!boost_asio_detail_posix_thread_function(+0x13)
0x00007f7082cba609 libpthread.so.0!UnknownFunction(0x9608)
0x00007f7081c2e293 libc.so.6!clone(+0x42)

[2021.12.14-19.24.25:604][968]LogExit: Executing StaticShutdownAfterError

The same code causing this issue on Linux works as intended on Windows.

[redbaron]

@mkieweg , looks like unhandled exception somewhere in our use of cpprest code (https://github.com/microsoft/cpprestsdk/blob/bfe34877914fd280512c97d41aaa99e88564fd8d/Release/include/pplx/pplxtasks.h#L952)

does it stop execution, without exiting, when you run it under debugger? If so maybe this->_M_stackTrace at the place where SIGTRAP was raised could show a bit more about the task with uncaught exception.

Also note to myself, this SO question might be relevant: https://stackoverflow.com/questions/38552098/c-rest-sdk-casablanca-sigtrap

[redbaron]

@mkieweg , would you be be able to post debug logs from Nakama server around the time crash happens? This will help us identify exact calls being made.

[mkieweg]

@mkieweg , would you be be able to post debug logs from Nakama server around the time crash happens? This will help us identify exact calls being made.

@redbaron I can totally do that. However, I won't have time for that before the end of the week.

[mkieweg]

@redbaron sorry for getting back to you just now. Unfortunately I did not get meaningful logs for this from our Nakama dev server and I can't reproduce it with an isolated instance locally. Right now I don't have the resources to look deeper into that matter, sorry.

[redbaron]

If you have a chance, try running under debugger and see if generated SIGTRAP will cause it to stop like if it was a breakpoint. Then inspecting this->_M_stackTrace can reveal something useful.

[lugehorsam]

@mkieweg we've reworked the SDK and this issue should now be resolved.