Skip to content

BuildOnConda.md

Latest commit

 

History

History
143 lines (82 loc) · 7.16 KB

BuildOnConda.md

File metadata and controls

143 lines (82 loc) · 7.16 KB

Build for Anaconda

Conda package can be generated to support more Linux platforms, and Windows. Conda package build has been tested successfully on local host.
In the future, parallel-preprocessor will be available on conda-forge channel for conda install -c condaforge parallel-preprocessor This conda package for parallel-preprocessor has both c++ binary and python wrapper installed, just like build deb package from source by cmake.

conda config --add channels conda-forge
conda config --set channel_priority strict
conda install occt
conda install parallel-preprocessor

conda recipe: meta.yaml and build scripts

The meta.yaml and build scripts have been provided in the recipe folder of this repository.

The package config file meta.yaml can be generated by a command, or just adapted from an existing recipe online, e.g. https://github.com/conda/conda-recipes/tree/master/libtiff Also note that the compilers used by conda-build can be specified using a conda_build_config.yaml.

The build scripts build.sh (for Unix-like OS) and bld.bat (windows batch file) have the install instruction (cmake is called just as compiling from source code.), install to a tmp prefix, then compressed into a zipped file, i.e. conda package.

build conda package from local machine

parallel-preprocessor depends on OpenCASCADE conda package, whose conda packaging configuration can be found at https://github.com/conda-forge/occt-feedstock/. To build conda package from local machine, user should install all the necessary dependency, such as OpenCASCADE 7.x, TBB, if conda build failed to download the dependencies specified in meta.yaml.

To build conda package, first of all, install anaconda python3 and install necessary packages like conda-build, and run the command in the recipe folder of this repository: conda build meta.ymal

The conda build task is conducted a seperate virtual environment automatically, so is the test(run) process.
D:\Software\anaconda3\envs\cf\conda-bld\ppp_1597347412425 On windows, cmake isntalled to the base conda environment can still detect occt package installed in the conda-build tmp conda environment.

NOTE: conda will not auto clean those tmp virtual env, user must manually remove the build virtual env.

anaconda upload /opt/anaconda3/conda-bld/linux-64/parallel-preprocessor-0.3-oct_2020.tar.bz2

To have conda build upload to anaconda.org automatically, use

conda config --set anaconda_upload yes

install built local conda package

Where is the generated conda package?
If testing script passed, the generated conda package, a compressed file, should be copied to the current working directory.
If test failed, there may be error message:

Tests failed for ppp-0.3-hdef9aff_2003.tar.bz2 - moving package to D:\Software\anaconda3\envs\cf\conda-bld\broken

After conda build, the built package can be installed: conda install --use-local ppp or equally conda install -c local ppp to install the built pkg. conda install --use-local ppp --dry-run to test if built package can be found and installed. conda verify may also verify the package without installation.

It is expected binary conda package for parallel-preprocessor will be hosted on github or conda-forge. After downloaded from github Release, the compressed package file can be installed by the command

conda install package-path/package-filename.tar.bz2

Conda build variants

https://docs.conda.io/projects/conda-build/en/latest/resources/variants.html

selector document

Here is an example to skip some variants in meta.yaml file.

build:
    skip: true  # [win or osx]

build for multiple python versions

Since python the optional interface module for parallel preprocessor is C-extension, so separate binary packages are needed for conda python 3.6, 3.7 and 3.8. Depends on platforms, conda python 3.8 may be minimum version to support C++17.

conda build support build for multiple version, by specify --py option multiple times. conda build --py 3.6 --py 3.7 --py 3.8 recipe if this does not work, try conda build recipe --variants "{'python': ['2.7', '3.6'], 'vc': ['9', '14']}"

After all, variant control is done by the yaml fileconda_build_config.yaml. If this file does not exists same directory as meta.yaml, conda build may generate one. real also Conda multi variant builds

conda-forge build environment

Build environments in conda-forge CI (miniconda azure pipeline, ) is quite different from conda build on local machine. After trials and referencing to https://github.com/FreeCAD/FreeCAD/blob/master/conda/, finally conda build is working.

conda windows python 3.8 using "VS 2017 update 9", although it is called vs2015_runtime indicating binary compatible, it may be using vs2017 (14.1) compiler.

vs2015_runtime 14.16.27012 MSVC runtimes associated with cl.exe version 19.16.27032.1 (VS 2017 update 9) / None
vc 14.1 A meta-package to impose mutual exclusivity among software built with different VS versions / Modified BSD License (3-clause)

The default conda macos conda-build using MacOS SDK 10.09, XCode 11.6, but MacOS 10.13 is needed for C++17 support. Therefore, the MacOS deploy target has been limited to "10.15", in <../recipe/conda_build_config.yaml>

/src/PropertyContainer/PropertyContainerTest.cpp:46:22: error: 'any_cast<std::__1::shared_ptr >' is unavailable: introduced in macOS 10.13 auto data = std::any_cast<std::shared_ptr>(a);

libGL is needed to build OCCT and parallel-preprocessor.

Upload to conda forge (yet done)

Here is guide from conda-forge website:

If your package is not on conda-forge yet you can follow their contribution documentation here: https://conda-forge.org/#add_recipe

The process: https://conda-forge.org/docs/maintainer/adding_pkgs.html

  1. fork conda-forge example repo

https://github.com/qingfengxia/staged-recipes

  1. push it as a new repo from the branch "parallel-preprocessor"

conda-forge/staged-recipes#12901 (comment)

get attention from gitter https://gitter.im/conda-forge/conda-forge.github.io

  1. later the maintainer (pusher) will have the write control on that repo.

Conda recipe feedback

chrisburr Contributor

I guess occt is being used as a shared library? If so it should use pin_compatible to make sure an ABI compatible version is used.

Author:

yes, occt is build by one FreeCAD developer. it should be something like numpy version, must be matched. shall I use occt == 7.4? I propabaly needs to follow the update of occt version, somehow, I know occt 7.4 is the only version on conda-forge

tbb also has the API compatibility issue