11:01 <ctismer> Hi cfbolz et al, I got a funny error using threads.

11:02 <ctismer> https://www.irccloud.com/pastebin/lrUB4U8T/

11:02 <ctismer> RUN 1: (For PyPy, see https://bitbucket.org/pypy/pypy/issues/2274)

11:02 <ctismer> Fatal RPython error: a thread is trying to wait for the GIL, but the GIL was not initialized

11:03 <ctismer> cfbolz: converting that link to the new website, I saw that this issue is 5 years old and should be solved.

11:08 <cfbolz> ctismer: ugh, no idea

11:08 <cfbolz> ctismer: any weird things you do with threads?

11:09 <ctismer> nothing, this is our standard thread implementation. (which can be weird enough, no idea :) )

11:11 <ctismer> cfbolz: I'll probably fire up the soldering iron and debug Windows :(

11:15 <cfbolz> ctismer: but it is not some kind of embedding scenario, right?

11:15 <cfbolz> are you using fork?

11:18 <ctismer> cfbolz: not at all. For instance, we get this message when running bug_462.py

11:18 <ctismer> cfbolz https://www.irccloud.com/pastebin/1yHQrD68/bug_462.py

11:18 <cfbolz> ctismer: do you make any threads from C, and then call python code from them?

11:19 <cfbolz> it's entirely possible that there is a bug in those parts of the C API emulation

11:19 <cfbolz> I suppose QThread starts threads from C code

11:20 <ctismer> cfbolz: I have to study the code then.

11:21 <ctismer> (we are about to blog about PyPy and tested things on Windows)

11:23 <cfbolz> right

11:26 <cfbolz> ctismer: I don't really see a test that makes a thread from C code and then call PyGILState_Ensure or something like that

11:26 <cfbolz> so I suppose that could be just broken, potentially

11:26 <cfbolz> but we are pretty far outside of my comfort zone here

11:28 <ctismer> cfbolz: you mean you love threads as much as I do? 😂

11:28 <cfbolz> exactly :-)

12:37 <cfbolz> ctismer: yeah, my suspicion remains that this is just not working with cpyext yet

12:38 <cfbolz> cffi has explicit tests around the problem

12:39 <cfbolz> we would have to write a small C extension that starts a new thread from C, then calls python

12:39 <cfbolz> to see whether it crashes

12:52 <ctismer> cfbolz: but why on earth does it work on macOS? I would have understood Windows, but not Linux

12:53 <cfbolz> yes, that is pretty weird 😆

15:52 <ctismer> cfbolz: I tested myself on my virtual Windows machine on the Mac - the effect does _not_ show up. It happens on a real Windows and Ubuntu machine in Berlin - even weirder cause I cannot solve it from here.

15:53 <cfbolz> 😞