Loading PyCall causes segfault on any subsequent Julia interrupt

[marius311]

On this one particular system I'm using, whenever I load PyCall, any time after I try to Ctrl+C interrupt, I get a segfault. Here's e.g. a single session, the first interrupt is fine, then after loading PyCall its a segfault.

julia> inv(rand(3000,3000))
^CERROR: InterruptException:
Stacktrace:
 [1] Array
   @ ./boot.jl:448 [inlined]
 [2] getri!(A::Matrix{Float64}, ipiv::Vector{Int64})
   @ LinearAlgebra.LAPACK /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.6/LinearAlgebra/src/lapack.jl:1043
 [3] inv!
   @ /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.6/LinearAlgebra/src/lu.jl:475 [inlined]
 [4] inv(A::Matrix{Float64})
   @ LinearAlgebra /buildworker/worker/package_linux64/build/usr/share/julia/stdlib/v1.6/LinearAlgebra/src/dense.jl:812
 [5] top-level scope
   @ REPL[1]:1

julia> using PyCall

julia> inv(rand(3000,3000))
^CFatal Python error: Segmentation fault

Current thread 0x00002aaaaab3f380 (most recent call first):

signal (11): Segmentation fault
in expression starting at none:0
raise at /lib64/libpthread.so.0 (unknown line)
_IO_funlockfile at /lib64/libpthread.so.0 (unknown line)
maybe_collect at /buildworker/worker/package_linux64/build/src/julia_threads.h:313 [inlined]
jl_gc_pool_alloc at /buildworker/worker/package_linux64/build/src/gc.c:1176
push! at ./array.jl:936 [inlined]

This is 100% reproducible, and it segfaults 100% of the time after loading PyCall. This is with PyCall 1.92.2, Julia 1.6-beta1 installed from binaries with 1 thread and Python 3.7.3 (default, Mar 27 2019, 22:11:17) [GCC 7.3.0] :: Anaconda, Inc. on linux.

Any suggestions for what I can do to try and debug this further / fix it?

[marius311]

I've got an even simpler reproducer. I'm guessing its the same issue, the stack trace is identical and its also triggered by loading PyCall. Run the following with julia --threads=2:

using PyCall

Threads.@sync begin
    Threads.@spawn GC.gc(true)
    Threads.@spawn GC.gc(true)
end

Commenting out the using PyCall removes the segfault.

[stevengj]

Closing in favor of #882.