Print crashes when using Nemo from another module

[Mao Zeng]

I'm trying to use Nemo to access Flint's multivariate polynomials, but run into a strange problem. A toy example is this. Here's a file "test1.jl":

# test1.jl

module test1

export x,y

using Nemo

Ring, (x,y) = PolynomialRing(ZZ, ["x", "y"])

end

Now I run these line in the Julia REPL:

push!(LOAD_PATH, ".") # add current directory to package search path

using test1

println(x + y)

The last line results in a segmentation fault. However, the following lines run without problem in the Julia REPL:

include("test1.jl")

println(test1.x+test1.y)

# print output: x + y

This workaround is inconvenient, since I want the variables / namespace to be exported.

I don't know if this is specific to Nemo, or it's because of my ignorance about using Julia modules in a nested manner.

[tho...@gmail.com]

Could you please provide `versioninfo(verbose = true)` and Pkg.status()?

[Mao Zeng]

julia> versioninfo(verbose=true)

Julia Version 1.5.3

Commit 788b2c77c1 (2020-11-09 13:37 UTC)

Platform Info:

  OS: Linux (x86_64-pc-linux-gnu)

      Ubuntu 20.04.2 LTS

  uname: Linux 5.4.0-65-generic #73-Ubuntu SMP Mon Jan 18 17:25:17 UTC 2021 x86_64 x86_64

  CPU: Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz: 

              speed         user         nice          sys         idle          irq

       #1  2719 MHz     336487 s       1001 s      84220 s   18617291 s          0 s

       #2  2545 MHz     333565 s       2495 s      84938 s   18615669 s          0 s

       #3  2560 MHz     332332 s       1834 s      91239 s   18604872 s          0 s

       #4  2550 MHz     326774 s       1712 s      86333 s   18601649 s          0 s

       #5  2619 MHz     334680 s       2238 s      83019 s   18602381 s          0 s

       #6  2759 MHz     333976 s       3927 s      82765 s   18610945 s          0 s

       #7  2681 MHz     314221 s       1362 s      91307 s   18629689 s          0 s

       #8  2698 MHz     334525 s       5269 s      89576 s   18610551 s          0 s

       

  Memory: 15.402080535888672 GB (240.1640625 MB free)

  Uptime: 190905.0 sec

  Load Avg:  1.35205078125  1.2802734375  1.34423828125

  WORD_SIZE: 64

  LIBM: libopenlibm

  LLVM: libLLVM-9.0.1 (ORCJIT, skylake)

Environment:

  MATHEMATICA_HOME = /usr/local/Wolfram/Mathematica/12.1

  WINDOWPATH = 2

  HOME = /home/mzeng

  PYTHONPATH = /home/mzeng/code/pySecDec-1.4.3/pySecDec-1.4.3:

  TERM = xterm-256color

  PATH = /home/mzeng/.cargo/bin:/home/mzeng/.local/bin:/home/mzeng/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/usr/games:/usr/local/games:/snap/bin

####################

julia> Pkg.status()

Status `~/.julia/environments/v1.5/Project.toml`

  [c3fe647b] AbstractAlgebra v0.12.0

  [861a8166] Combinatorics v1.0.2

  [864edb3b] DataStructures v0.18.9

  [c27321d9] Glob v1.3.0

  [7073ff75] IJulia v1.23.1

  [50d2b5c4] Lazy v0.15.1

  [18c93696] MathLink v0.3.1

  [3b2b4ff1] Multisets v0.3.3

  [2edaba10] Nemo v0.19.1

  [c46f51b8] ProfileView v0.6.9

  [123dc426] SymEngine v0.8.3

  [ac1d9e8a] ThreadsX v0.1.7

[Mao Zeng]

By the way, are you able to reproduce the problem as I initially reported? Or do you think it's related to the specific OS or Julia version I'm using?

[tho...@gmail.com]

OK, I remember what is going on. This is https://github.com/Nemocas/Nemo.jl/issues/790. It is a problem with precompilation. Here is the solution:

function __init__()

  global Ring, (x,y) = PolynomialRing(ZZ, ["x", "y"])

end

[Mao Zeng]

Thanks, that solved the problem! Using "__init__" is even necessary when I access x and y from within the module, like this

# new version of test1.jl

module test1

export print1

using Nemo

function __init__()

   global Ring, (x,y) = PolynomialRing(ZZ, ["x", "y"])

end

print1() = println(x+y)

end

After loading the above module, running print1() would trigger a segmentation fault if I didn't use the "__init()__". I'm still mystified by what Julia is doing internally for precompilation, but at least the problem was solved.

[tho...@gmail.com]

You can have a look here if you are interested in the details: https://github.com/JuliaLang/julia/issues/15722