Have you ever seen a package named jll, like Arpack_jll, when using Julia? Most of the time, when you install some package, you will probably have a package with jll as a dependency. In fact, this package with jll is a compilation of code written in a language other than Julia. When installing, the binary will be downloaded automatically according to the OS.
This is achieved BinaryBuilder.jl https://github.com/JuliaPackaging/BinaryBuilder.jl It is a package called.
I https://www.youtube.com/watch?v=3IyXsBwqll8 I knew the existence by seeing this.
Apparently, Docker is used to cross-compile and compile code on various platforms of Mac, Windows, and Linux. At that time, if it does not work for each OS, it seems that the compilation method is modified as appropriate.
It's a very interesting library, so I tried it.
This time I put in C-LIME. https://usqcd-software.github.io/c-lime/ A data format for lattice QCD (lattice quantum chromodynamics). I'm not familiar with it at all, but the scale was just right so I tried it.
Julia 1.5.2 Mac OS 10.15
First, you need Docker, so put Docker in it. Please find out how to install Docker. For Mac
$ brew install docker
$ brew cask install docker
It looks like it can be done.
Then in Julia, press] to enter package mode,
add BinaryBuilder
will do. However, as of October 17, 2020, something has changed on GitHub that doesn't work with the current version, so instead of the above
add BinaryBuilder#f126fe7
You need to put in the modified version as.
After putting it in
using BinaryBuilder
BinaryBuilder.run_wizard()
And run wizard.
Make a platform selection
> All Supported Platforms
Select by Operating System
Fully Custom Platform Choice
Then select All Supported Platforms and
Please enter a URL (git repository or compressed archive) containing the source code to build:
Now specify the github address or file of the software you want to compile. here,
https://github.com/usqcd-software/c-lime.git
It was made. Then follow the instructions. The first question is which of the git branches to choose, but let's leave this as master. This code does not require any additional source code or libraries. As you proceed, There are things like giving a name and giving a version, so I will put it in as appropriate.
As you proceed
sandbox:${WORKSPACE}/srcdir
Something like that will come out. Now compile the code. In this case,
cd c-lime/
./autogen.sh
./configure --prefix=${prefix} --build=${MACHTYPE} --host=${target}
make
make install
You can compile the code with. Then press Ctrl + D to log out. next
Your build script was:
cd $WORKSPACE/srcdir
cd c-lime/
./autogen.sh
./configure --prefix=${prefix} --build=${MACHTYPE} --host=${target}
make
make install
Would you like to edit this script now? [y/N]: n
Since the compilation method is summarized, set it to n if there is no particular problem. Then what do you compile? Something like that will appear, so press a and then d. After that, compile on various OSs and platforms while pushing enter.
Well, after it's done,
How should we deploy this build recipe?
> Prepare a pull request against the community buildtree, Yggdrasil
Write to a local file
Print to stdout
Will be asked. If you choose the first, you will make a pull request to Yggdrasil on GitHub. This means putting information here on how to compile.
Would you like to actually open the pull request or just prepare it?
> Go ahead, open it
No, just prepare it and let me look at it first
After that, you will make a pull request with your github account name.
If it finishes successfully, a pull request will be made automatically. And if there is no problem, it will be merged into Yggdrasil. Once merged, https://github.com/JuliaBinaryWrappers/ Will be registered in. If you register here, you're done. https://github.com/JuliaBinaryWrappers/CLIME_jll.jl Is already registered,
Linux aarch64 {libc=glibc} (aarch64-linux-gnu)
Linux aarch64 {libc=musl} (aarch64-linux-musl)
Linux armv7l {call_abi=eabihf, libc=glibc} (armv7l-linux-gnueabihf)
Linux armv7l {call_abi=eabihf, libc=musl} (armv7l-linux-musleabihf)
Linux i686 {libc=glibc} (i686-linux-gnu)
Linux i686 {libc=musl} (i686-linux-musl)
Windows i686 (i686-w64-mingw32)
Linux powerpc64le {libc=glibc} (powerpc64le-linux-gnu)
macOS x86_64 (x86_64-apple-darwin)
Linux x86_64 {libc=glibc} (x86_64-linux-gnu)
Linux x86_64 {libc=musl} (x86_64-linux-musl)
FreeBSD x86_64 (x86_64-unknown-freebsd)
Windows x86_64 (x86_64-w64-mingw32)
It is said that the platform can provide binaries.
After that
add CLIME_jll
Then you can install it. This CLIME is not a library but a type to execute, so https://juliapackaging.github.io/BinaryBuilder.jl/dev/jll/ With reference to
using CLIME_jll
lime_contents() do exe
run(`$exe`)
end
By doing so, you can execute the executable file lime_contents on Julia.
Note that this code actually requires arguments, so use it in the form of `run (` $ exe lime_file_test`) `.
This time I ran BinaryBuilder.jl on a Mac, but that alone created a Julia library for various platforms. I'm surprised.