Install scipy dependency from pypi (#202)

* Install scipy via pip if not installed

* Remove LD_PRELOAD from CI

* Pip install scipy during build phase

* Don't install deps

* Try removing scipy before installing

* Remove undefined keyword

* Explicitly build package before docs

.github/workflows/ci.yml

@@ -23,8 +23,6 @@ jobs:
         os: [ubuntu-latest]
     steps:
       - uses: actions/checkout@v2
-      - name: Set LD_PRELOAD
-        run: echo "LD_PRELOAD=$HOME/.julia/conda/3/lib/libstdc++.so" >> $GITHUB_ENV
       - uses: julia-actions/setup-julia@v1
         with:
           version: ${{ matrix.julia-version }}

.github/workflows/documenter.yml

@@ -17,8 +17,6 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v2
-      - name: Set LD_PRELOAD
-        run: echo "LD_PRELOAD=$HOME/.julia/conda/3/lib/libstdc++.so" >> $GITHUB_ENV
       - uses: julia-actions/setup-julia@v1
       - name: Cache CmdStan
         id: cache-cmdstan
@@ -33,6 +31,7 @@ jobs:
           tar -xzpf ${{ env.CMDSTAN_PATH }}/cmdstan-${{ env.CMDSTAN_VERSION }}.tar.gz -C ${{ env.CMDSTAN_PATH }}
           make -C ${{ env.CMDSTAN_PATH }}/cmdstan-${{ env.CMDSTAN_VERSION }}/ build
         shell: bash
+      - uses: julia-actions/julia-buildpkg@latest
       - uses: julia-actions/julia-docdeploy@v1
         env:
           JULIA_CMDSTAN_HOME: ${{ env.CMDSTAN_PATH }}/cmdstan-${{ env.CMDSTAN_VERSION }}/ # required by CmdStan.jl

.github/workflows/futures.yml

@@ -28,8 +28,6 @@ jobs:
             arviz_version: "release"
     steps:
       - uses: actions/checkout@v2
-      - name: Set LD_PRELOAD
-        run: echo "LD_PRELOAD=$HOME/.julia/conda/3/lib/libstdc++.so" >> $GITHUB_ENV
       - uses: julia-actions/setup-julia@v1
         with:
           version: ${{ matrix.julia-version }}

.gitignore

@@ -2,6 +2,7 @@
 *.jl.*.cov
 *.jl.mem
 /deps/deps.jl
+/deps/build.log
 /docs/src/quickstart.md
 /docs/src/creating_custom_plots.md
 /docs/src/*.log

deps/build.jl

@@ -0,0 +1,7 @@
+using Conda
+
+# try to install scipy with pip if not yet installed
+# temporary workaround for https://github.com/arviz-devs/ArviZ.jl/issues/188
+Conda.pip_interop(true)
+Conda.pip("uninstall -y", "scipy")
+Conda.pip("install", "scipy")