Thread race condition issue when running with MSYS2 on Windows

Great program! Works fine on Ubuntu, but not on Windows.

I get segfault doing anything with it built with Msys2 MinGW64 on Windows.

$ gdb system.exe
GNU gdb (GDB) 15.2
Copyright (C) 2024 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <[http://gnu.org/licenses/gpl.html>](http://gnu.org/licenses/gpl.html%3E)
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Type "show copying" and "show warranty" for details.
This GDB was configured as "x86_64-w64-mingw32".
Type "show configuration" for configuration details.
For bug reporting instructions, please see:
<[https://www.gnu.org/software/gdb/bugs/>](https://www.gnu.org/software/gdb/bugs/%3E).
Find the GDB manual and other documentation resources online at:
    <[http://www.gnu.org/software/gdb/documentation/>](http://www.gnu.org/software/gdb/documentation/%3E).
For help, type "help".
Type "apropos word" to search for commands related to "word"...
Reading symbols from system.exe...
(gdb) run test 1 2
Starting program: D:\projects\Mahjong\akochan\system.exe test 1 2
[New Thread 57400.0xe570]
[New Thread 57400.0x674]
[New Thread 57400.0x2200]
0 {"aka_flag": true, "kyoku_first": 4, "type": "start_game"}
1 {"bakaze": "E", "dora_marker": "3m", "haiyama": ["7s", "6p", "6m", "9m", "E", "2m", "S", "2p", "9s", "6m", "1s", "N", "5p", "2p", "4m", "2s", "1m", "2p", "1s", "S", "2p", "S", "4p", "5mr", "3p", "4s", "C", "5sr", "E", "5p", "E", "6p", "6p", "1s", "E", "3s", "8s", "7p", "P", "2s", "5s", "8s", "9m", "8p", "6m", "4m", "5pr", "5m", "8m", "6s", "5m", "9m", "P", "F", "2m", "7p", "6m", "7p", "3p", "3s", "7m", "2s", "4p", "7s", "6s", "5m", "3p", "8m", "3p", "4s", "C", "1m", "N", "F", "8p", "1p", "4m", "1p", "1p", "W", "P", "1s", "9s", "5p", "6p", "P", "9p", "W", "9s", "4m", "8s", "C", "3s", "3m", "S", "W", "1m", "4p", "9s", "1m", "8p", "6s", "W", "9p", "5s", "7s", "8p", "4s", "7m", "N", "9p", "7m", "2s", "6s", "2m", "1p", "7s", "N", "8m", "3m", "2m", "7m", "C", "F", "F", "5s", "8s", "9p", "3m", "4s", "3m", "8m", "3s", "4p", "9m", "7p"], "honba": 0, "kyoku": 1, "kyotaku": 0, "oya": 0, "scores": [25000, 25000, 25000, 25000], "tehais": [["7s", "E", "9s", "5p", "1m", "2p", "3p", "E", "6p", "8s", "5s", "6m", "8m"], ["6p", "2m", "6m", "2p", "2p", "S", "4s", "5p", "1s", "7p", "8s", "4m", "6s"], ["6m", "S", "1s", "4m", "1s", "4p", "C", "E", "E", "P", "9m", "5pr", "5m"], ["9m", "2p", "N", "2s", "S", "5mr", "5sr", "6p", "3s", "2s", "8p", "5m", "9m"]], "type": "start_kyoku"}
2 {"actor": 0, "pai": "P", "type": "tsumo"}
[New Thread 57400.0xed98]
[New Thread 57400.0xe70]
[New Thread 57400.0x4928]
[New Thread 57400.0x7104]
[New Thread 57400.0x4204]
[New Thread 57400.0xef00]
[New Thread 57400.0xe2c8]
[New Thread 57400.0x9848]
[New Thread 57400.0xecb0]
[New Thread 57400.0x1fa8]
[New Thread 57400.0x8188]
[New Thread 57400.0xd4a0]
[New Thread 57400.0xaad4]
[New Thread 57400.0xe180]
[New Thread 57400.0xcd30]
[New Thread 57400.0xbd04]
[New Thread 57400.0xe4d8]
[New Thread 57400.0xeca4]
[New Thread 57400.0x37c8]
[New Thread 57400.0xc5b8]
[New Thread 57400.0x79f8]
[New Thread 57400.0x700]
[New Thread 57400.0xa928]
Thread 1 received signal SIGSEGV, Segmentation fault.
Tehai_Calculator::set_candidates3_multi_thread (this=0x5a95f0, game_record=..., game_state=...,
    sute_flag=..., tehai_analyzer=..., tactics=...) at tehai_cal.cpp:887
887                                     const Agari_Calc& agari = tehai_calculator_work.agari_graph_work[agari_loc[0]][an];
(gdb)

Anyway, this problem can be worked around by limiting to one thread:

export OMP_NUM_THREADS=1
./system.exe test 1 2

(works fine)

Random thoughts:

Linux's native threading and OpenMP implementation is generally more robust than Windows/MSYS2
tehai_calculator_work is a global variable shared across threads, the thread safety is probably just not happening with this build for whatever reason

critter-mj / akochan

Thread race condition issue when running with MSYS2 on Windows #27