Windows: sclang crashes on executing menu-item "Quit interpreter"/freezes on evaluating 0.exit

[bagong]

Studied on current scwin3.7 alpha01, Visual Studio build, the problem has been around since the rebirth of the 3.7 Windows version.
When running "Quit Interpreter" from the IDE-Language menu, sclang will quit, but it actually crashes, Postwindow:

Interpreter has crashed or stopped forcefully. [Exit code: 62097]

The menu point calls stopLanguage() in sc_process.cpp which in turn evaluates 0.exit. When that fails (not succeeded after 200 ms) it kills sclang (according to Google, Exit code: 62097 indicates that that happened).

Running 0.exit in the editor gets stuck on main: waiting for input thread to join...

SURPRISE: if I press the key Enter in this state, the shutdown continues and I get this post:

main: quitting...
empty
cleaning up OSC

And when the server is running while this happens, the server is shut down:

main: quitting...
/quit sent

empty
cleaning up OSC

So server not quitting on IDE-exit a side-effect of of sclang getting stuck during shutdown before the server is quit, and then getting killed.

When closing sclang with 0.exit and kicking off the process by hitting Enter, as said, the server is quit as expected, but after that a(nother) sclang-crash still occures, in this case the exit code is -1073740791

There are various issues of sclang borking on close. In Linux currently sclang also crashes on exit, but in a less severe way, the server is closed and the two main-steps and the server-shutdown occure before the crash happens, could be the same as the second crash on Windows.

[bagong]

#1449

[bagong]

#1422 #939 #1438

[bagong]

In the mingw/msys2 32-bit build there are subtle differences.

0.exit also requires a Enter to complete, but then it doesn't crash, it posts:

Interpreter has quit. [Exit code: 3]

If the server is running, it will be shut down by this procedure, but then the IDE goes in a frozen state and the sclang shutdown doesn succeed. No return message after "cleaning up OSC". So that second sclang crash seems to have severe consequences in the mingw build

[bagong]

If I run 0.exit in the IDE in a debug build, nothing happens until I press enter. Then this message window pops up:
Debug Error! Program .....\sclang.exe
R6010

• abort() has been called
  Press retry to debug application

Then it is offered to create a breakpoint in a window headed DiskIO_UGens/std::thread/~thread() at
~thread() _NOEXCEPT
{ // cleanup
if (joinable()) _XSTD terminate()}
}

If I continue, the interpreter quits with exit code 3

The server cannot be started in debug mode, so I cannot produce any output with a running server.

I'll have to leave it at that, unfortunately I cannot interpret these informations.

[timblechmann]

306c417?

btw, these crash data re not helpful: install a debugger (e.g. windows platform sdk), attach to the sclang or scsynth process, crash the program, post the stack trace

[telephon]

Added a note in known issues

[llloret]

I think this is fixed with 3.7.2.

[bagong]

works in 3.8 too, thank you Luis!