(Unintended) Destruction of Thread object causes hard to debug crash

[skunkiferous]

Me, and people trying to help me on the forum, spent a while to debug a crash, which was due to a Thread object being destroyed, while the thread is still running. In this particular case, I believe the Thread object was on the stack, but I assume the same would happen with a GCed Thread object.

Here is a simple example that causes the crash, at least for me on Windows, and was also reproducible on Archlinux:

import os

proc whatever() {.thread, nimcall.} =
  echo("TEST")

proc initProcess(): void =
  echo("In initProcess()")
  var thread: Thread[void]
  createThread(thread, whatever)
  echo("initProcess() done")

proc doStuff(): void =
  echo("In doStuff()")
  # ...
  initProcess()
  sleep(500)
  # ...
  echo("Crashes before getting here!")

doStuff()

If "var thread" is moved to the global scope, the crash does not happen.

It is not clear to me if the crash can be prevented though a compile-time error, but at least it would be good to have an error message that users understand. It might also be good to mention that this can happen in the Thread and/or createThread documentation.

[cheatfate]

@andreaferretti, please unmark Threads, because this is not bug of threads. If people creates something in stack, they need to be ready that this will be lost after procedure exit.

[andreaferretti]

I agree it is not a bug of threads - in some sense it is not a bug at all. Still I thought it would be useful to locate issues such as this when looking for thread-related problems

[ghost]

Doesnt crash for me. Nim latest head.

[ghost]

Crashes for me on Arch Linux with Clang and GCC
@konqoro what is your platform/compiler?

[cheatfate]

@konqoro, please run executable many times , crashes can be not constant...

[ghost]

Tried again and without '-d:release' crashes every single time. with release doesn't. gcc version 7.2.1 20180116 (GCC) manjaro

[genotrance]

Confirmed works on Windows, crashes on Ubuntu.

[narimiran]

without '-d:release' crashes every single time

I've run it several times (without -d:release) on Manjaro Linux, with Nim 1.5, and it never crashed.

[kobi2187]

I think that it is not clear from the docs what happens on createThread.
If the thread instance is added to some global sequence, and can be retrieved later, then I guess it shouldn't crash. But if you're trying to run an operation through an object which by this time is already nil, you will get a runtime error, in other words, a crash. Should the compiler warn? usually it doesn't know about runtime state.