Can't install MPI.jl on Windows

[tadeu]

I'm getting these errors:

julia> Pkg.add("MPI")
INFO: Cloning cache of BinDeps from git://github.com/JuliaLang/BinDeps.jl.git
INFO: Cloning cache of Compat from git://github.com/JuliaLang/Compat.jl.git
INFO: Cloning cache of MPI from git://github.com/JuliaParallel/MPI.jl.git
INFO: Cloning cache of SHA from git://github.com/staticfloat/SHA.jl.git
INFO: Cloning cache of URIParser from git://github.com/JuliaWeb/URIParser.jl.git
INFO: Installing BinDeps v0.3.21
INFO: Installing Compat v0.7.14
INFO: Installing MPI v0.4.0
INFO: Installing SHA v0.1.2
INFO: Installing URIParser v0.1.3
INFO: Building MPI
=================================[ ERROR: MPI ]=================================

LoadError: None of the selected providers can install dependency juliampi.
Use BinDeps.debug(package_name) to see available providers

while loading D:\9\.julia\v0.4\MPI\deps\build.jl, in expression starting on line 54

================================================================================

==========================================[ BUILD ERRORS ]==========================================


WARNING: MPI had build errors.

 - packages with build errors remain installed in D:\9\.julia\v0.4
 - build the package(s) and all dependencies with `Pkg.build("MPI")`
 - build a single package by running its `deps/build.jl` script

====================================================================================================

INFO: Package database updated

julia> import BinDeps

julia> BinDeps.debug("MPI")
INFO: Reading build script...
The package declares 1 dependencies.
 - Library "juliampi"
    - Providers:
      - Simple Build Process (can't provide)

What should I do?

[tadeu]

Ok, I tried to add a Windows provider that uses cmake and ninja, but yet, it won't work:

provides(SimpleBuild,
    (@build_steps begin
        CreateDirectory(builddir)
        (@build_steps begin
            ChangeDirectory(builddir)
            (@build_steps begin
                configure
                `ninja`
                `ninja install`
            end)
        end)
    end), [mpi], os = :Windows)

it just says:

julia> Pkg.build("MPI")
INFO: Building MPI
===========================================[ ERROR: MPI ]===========================================


LoadError: None of the selected providers can install dependency juliampi.
Use BinDeps.debug(package_name) to see available providers

while loading D:\9\.julia\v0.5\MPI\deps\build.jl, in expression starting on line 67

====================================================================================================


==========================================[ BUILD ERRORS ]==========================================


WARNING: MPI had build errors.

 - packages with build errors remain installed in D:\9\.julia\v0.5
 - build the package(s) and all dependencies with `Pkg.build("MPI")`
 - build a single package by running its `deps/build.jl` script

====================================================================================================

Can I make it run verbosely, to see what error is happening?

[davidanthoff]

master has support for MS MPI, should be enough to have MS MPI installed.

[blegat]

This is because by default, if you are on Windows, BinDeps does not try BuildPropcess providers.
Since SimpleBuild is a BuildProcess, BinDeps does not find any providers.
You can check here, you see the lines

# Default installation method
if OS_NAME == :Darwin
    defaults = [Binaries,PackageManager,SystemPaths,BuildProcess]
elseif OS_NAME == :Linux
    defaults = [PackageManager,SystemPaths,BuildProcess]
elseif OS_NAME == :Windows
    defaults = [Binaries,PackageManager,SystemPaths]
else
    defaults = [SystemPaths,BuildProcess]
end

This is discussed in this issue.
You can add BuildProcess for Windows by changing the end of MPI/deps/build.jl into

@windows_only push!(BinDeps.defaults, BuildProcess)

@BinDeps.install Dict(:mpi => :mpi)

@windows_only pop!(BinDeps.defaults)

[simonbyrne]

#271 got rid of BinDeps, so this should be fixed. If you have more problems, please open a new issue.