Misleading build error on unsupported architecture

[omus]

When attempting to build PyCall on Sys.ARCH of :aarch64 with the environmental variable PYTHON="" set you get the following error message:

┌ Error: Error building `PyCall`:
│ ERROR: LoadError: No system-wide Python was found; got the following error:
│ UseCondaPython()
│ Stacktrace:
│  [1] error(::String, ::String) at ./error.jl:42
│  [2] top-level scope at /root/.julia/packages/PyCall/zqDXB/deps/build.jl:77
│  [3] include at ./boot.jl:328 [inlined]
│  [4] include_relative(::Module, ::String) at ./loading.jl:1105
│  [5] include(::Module, ::String) at ./Base.jl:31
│  [6] include(::String) at ./client.jl:424
│  [7] top-level scope at none:5
│ in expression starting at /root/.julia/packages/PyCall/zqDXB/deps/build.jl:43
│ caused by [exception 1]
│ UseCondaPython()
│ Stacktrace:
│  [1] top-level scope at /root/.julia/packages/PyCall/zqDXB/deps/build.jl:51
│  [2] include at ./boot.jl:328 [inlined]
│  [3] include_relative(::Module, ::String) at ./loading.jl:1105
│  [4] include(::Module, ::String) at ./Base.jl:31
│  [5] include(::String) at ./client.jl:424
│  [6] top-level scope at none:5
└ @ Pkg.Operations /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.3/Pkg/src/backwards_compatible_isolation.jl:649

If I add :aarch64 into the list of supported architectures I get a much better error message:

┌ Error: Error building `PyCall`:
│ ┌ Info: Using the Python distribution in the Conda package by default.
│ └ To use a different Python version, set ENV["PYTHON"]="pythoncommand" and re-run Pkg.build("PyCall").
│ [ Info: Downloading miniconda installer ...
│ ERROR: LoadError: Unsupported architecture: aarch64
│ Stacktrace:
│  [1] error(::String) at ./error.jl:33
│  [2] _installer_url() at /root/.julia/packages/Conda/3rPhK/src/Conda.jl:143
│  [3] _install_conda(::String, ::Bool) at /root/.julia/packages/Conda/3rPhK/src/Conda.jl:164
│  [4] _install_conda(::String) at /root/.julia/packages/Conda/3rPhK/src/Conda.jl:155
│  [5] runconda(::Cmd, ::String) at /root/.julia/packages/Conda/3rPhK/src/Conda.jl:112
│  [6] #add#1(::String, ::typeof(Conda.add), ::String, ::String) at /root/.julia/packages/Conda/3rPhK/src/Conda.jl:188
│  [7] add at /root/.julia/packages/Conda/3rPhK/src/Conda.jl:187 [inlined] (repeats 2 times)
│  [8] top-level scope at /root/.julia/packages/PyCall/gqZkc/deps/build.jl:86
│  [9] include at ./boot.jl:328 [inlined]
│  [10] include_relative(::Module, ::String) at ./loading.jl:1105
│  [11] include(::Module, ::String) at ./Base.jl:31
│  [12] include(::String) at ./client.jl:424
│  [13] top-level scope at none:5
│ in expression starting at /root/.julia/packages/PyCall/gqZkc/deps/build.jl:45
└ @ Pkg.Operations /buildworker/worker/package_linuxaarch64/build/usr/share/julia/stdlib/v1.3/Pkg/src/backwards_compatible_isolation.jl:649

Unsetting the PYTHON environmental variable allows the PyCall build to complete. It would be good to fix the PyCall error message so it isn't misleading.

[rapus95]

Given that Conda.jl supports miniforge I would even suggest to remove that architecture check. If Conda is selected, let Conda handle it.

[tpgillam]

I've also run into this problem just now.

@omus Out of interest, when you say:

Unsetting the PYTHON environmental variable allows the PyCall build to complete.

was this the only step that was required, or was the addition of :aarch64 somewhere also needed? For me just having PYTHON not set seems insufficient to get the build to complete. Thanks!

[omus]

@tpgillam I believe the commend of unsetting PYTHON was in addition to adding :aarch64 to the list of supported architectures in deps/build.jl.

If you just want to get PyCall working on aarch64 set:

export PYTHON=$(which python)

before starting Julia. You may need to run ] build PyCall with that environmental variable set.

[stevengj]

Given that Conda.jl supports miniforge I would even suggest to remove that architecture check. If Conda is selected, let Conda handle it.

I agree; see #906.

[tpgillam]

Thanks @omus! This does indeed work (along with installing a few additional dependencies for the system python — I was doing this inside a docker container running on an M1 mac).