SIGABRT when importing umap

[bicycle1885]

I'm trying to use the umap package via PyCall.jl. However, @pyimport umap results in assertion failure:

julia> using PyCall: @pyimport

julia> @pyimport umap
Assertion failed: (PassInf && "Expected all immutable passes to be initialized"), function addImmutablePass, file /Users/buildbot/miniconda3/conda-bld/llvmdev_1511828203076/work/lib/IR/LegacyPassManager.cpp, line 749.

signal (6): Abort trap: 6
while loading no file, in expression starting on line 467
__pthread_kill at /usr/lib/system/libsystem_kernel.dylib (unknown line)
Allocations: 5064700 (Pool: 5063238; Big: 1462); GC: 9
fish: 'julia' terminated by signal SIGABRT (Abort)

This is 100% reproducible on my machine, and never happens when importing it from Python3, so I think the problem is due to PyCall.jl or Julia.

I'm using PyCall.jl 1.15.0.

julia> versioninfo()
Julia Version 0.6.2
Commit d386e40c17 (2017-12-13 18:08 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin14.5.0)
  CPU: Intel(R) Core(TM) i5-6267U CPU @ 2.90GHz
  WORD_SIZE: 64
  BLAS: libopenblas (USE64BITINT DYNAMIC_ARCH NO_AFFINITY Haswell)
  LAPACK: libopenblas64_
  LIBM: libopenlibm
  LLVM: libLLVM-3.9.1 (ORCJIT, skylake)

julia> PyCall.pyversion
v"3.6.1"

julia> PyCall.libpython
"/Users/kenta/miniconda3/lib/libpython3.6m"

julia> PyCall.PYTHONHOME
"/users/kenta/miniconda3:/users/kenta/miniconda3"

~/.j/v/PyCall ((v1.15.0)|✔) $ pip list --format=columns | grep -E "numba|numpy|umap|scipy|scikit-learn"
numba                             0.37.0
numpy                             1.14.2
scikit-learn                      0.19.1
scipy                             1.0.0
umap-learn                        0.2.1

[stevengj]

It might be a conflict between the LLVM used by umap and the LLVM linked by Julia itself? See also #220

[stevengj]

See also JuliaLang/julia#12644 (comment) … a workaround is apparently to open the offending library with dlopen("...path...", Libdl.RTLD_DEEPBIND) for now.

[stevengj]

Closing as this needs to be fixed in Julia itself.