Heap corruption when printing, if importing pytorch *before* julia

[MilesCranmer]

I have no idea what this very strange issue comes from. It could be a combination of my specific torch version (tested 1.11-1.13, still occurs), specific operating system (macOS), specific python version (3.8.9) combined with my specific julia version (1.7.2).

Here's a minimal example:

import torch
from julia import Pkg
Pkg.add("StatsBase")

The output is:

Python(61891,0x100598580) malloc: *** error for object 0xffffffff00000000: pointer being freed was not allocated
Python(61891,0x100598580) malloc: *** set a breakpoint in malloc_error_break to debug

signal (6): Abort trap: 6
in expression starting at none:0
__pthread_kill at /usr/lib/system/libsystem_kernel.dylib (unknown line)
Allocations: 11431809 (Pool: 11423282; Big: 8527); GC: 14
[1]    61891 abort      ~/venvs/main/bin/ipython

However, I can make this go away with:

from julia import Pkg
import torch
Pkg.add("StatsBase")

This works fine.

I see this more helpful error when I run this code in a debugger:

File ~/venvs/main/lib/python3.8/site-packages/julia/core.py:621, in Julia.eval(self, src)
    619 if src is None:
    620     return None
--> 621 ans = self._call(src)
    622 if not ans:
    623     return None

File ~/venvs/main/lib/python3.8/site-packages/julia/core.py:549, in Julia._call(self, src)
    547 # logger.debug("_call(%s)", src)
    548 ans = self.api.jl_eval_string(src.encode('utf-8'))
--> 549 self.check_exception(src)
    551 return ans

File ~/venvs/main/lib/python3.8/site-packages/julia/core.py:603, in Julia.check_exception(self, src)
    601 else:
    602     exception = sprint(showerror, self._as_pyobj(res))
--> 603 raise JuliaError(u'Exception \'{}\' occurred while calling julia code:\n{}'
    604                  .format(exception, src))

JuliaError: Exception 'IOError: pipe_open: file already exists (EEXIST)' occurred while calling julia code:

So it seems like perhaps pytorch is taking over one of the pipes for printing from Python?