rkanadev / osxcross Goto Github PK

The goal of OSXCross is to provide a well working OS X cross toolchain for Linux, *BSD, and Cygwin.

Clang/LLVM is a cross compiler by default and is now available on nearly every Linux distribution, so we just need a proper port of the cctools (ld, lipo, ...) and the OS X SDK.

If you want, then you can build an up-to-date vanilla GCC as well.

Basically everything you can build on OS X with clang/gcc should build with this cross toolchain as well.

OSXCross comes with a minimalistic MacPorts Packet Manager. See README.MACPORTS for more.

Windows/Cygwin users should follow README.CYGWIN.

Move your packaged SDK to the tarballs/ directory.

Then ensure you have the following installed on your system:

Clang 3.2+, patch, libxml2-devel (<=10.6 only) and the bash shell.

Optional:

llvm-devel: For Link Time Optimization support uuid-devel: For ld64 -random_uuid support llvm-devel + xar-devel: For ld64 -bitcode_bundle support

You can find xar here. Do not install libxar-dev on Ubuntu, it's a different package.

-- You can run 'sudo tools/get_dependencies.sh' to get these automatically.

OSXCross uses clang as the default compiler for building its tools, and also as a cross-compiler to create OSX binaries.

In clang there is no difference between cross-compilation and native compilation, so OSXCross can use a normal clang install for both. You can use either a clang installation you already have, or build your own from source.

To build and install your own clang from a recent source tree, using gcc, run:

    ./build_clang.sh

This installs clang into /usr/local. If you want to install somewhere else, set the INSTALLPREFIX variable. For example:

    INSTALLPREFIX=/opt/clang ./build_clang.sh

On debian-like systems you can also use llvm.org/apt to get a newer version of clang. But be careful, that repository is known to cause troubles.

To build the cross toolchain (using clang), run:

    ./build.sh

Or, set variable UNATTENDED to 1 to skip the prompt and proceed straight to the build:

    UNATTENDED=1 ./build.sh

(This will search 'tarballs' for your SDK and then build in its own directory.)

Once this is done: add <path>/target/bin to your PATH variable so that you can invoke the cross-compiler.

That's it. See usage examples below.

If you also want to build GCC as a cross-compiler, you can do that by running:

    ./build_gcc.sh

The script lets you select a GCC version by setting the variable GCC_VERSION. By default you get C and C++ compilers, but you can tell the script to build a Fortran compiler as well:

    GCC_VERSION=5.2.0 ENABLE_FORTRAN=1 ./build_gcc.sh

[A gfortran usage example can be found here]

Before you do this, make sure you have the GCC build depedencies installed on your system.

On debian like systems you can install these using:

    sudo apt-get install gcc g++ zlib1g-dev libmpc-dev libmpfr-dev libgmp-dev

ATTENTION:

OSXCross links libgcc and libstdc++ statically by default (this affects -foc-use-gcc-libstdc++ too). You can turn this behavior off with OSXCROSS_GCC_NO_STATIC_RUNTIME=1 (env).

The build also creates aliases *-g++-libc++ which link with the clang implementation of the C++ standard library instead of the GCC version. Don't use these variants unless you know what you're doing.

Please ensure you have read and understood the Xcode license terms before continuing.

1. [Download Xcode **]
2. [Mount Xcode.dmg (Open With -> DiskImageMounter) ***]
3. Run: ./tools/gen_sdk_package.sh (from the OSXCross package)
4. Copy the packaged SDK (*.tar.* or *.pkg) on a USB Stick
5. (On Linux/BSD) Copy or move the SDK into the tarballs/ directory of OSXCross.

** Xcode up to 7.3.x is known to work.

*** If you get a dialog with a crossed circle, ignore it. You don't need to install Xcode.

Step 1. and 2. can be skipped if you have Xcode installed.

1. Download Xcode like described in 'Packaging the SDK on Mac OS X'
2. Install cmake, libxml2-dev and fuse
3. Run ./tools/gen_sdk_package_darling_dmg.sh <xcode>.dmg
4. Copy or move the SDK into the tarballs/ directory

1. Download Xcode like described in 'Packaging the SDK on Mac OS X'
2. Ensure you have clang and make installed
3. Run ./tools/gen_sdk_package_p7zip.sh <xcode>.dmg
4. Copy or move the SDK into the tarballs/ directory

1. Download Xcode 4.2 for Snow Leopard
2. Ensure you are downloading the "Snow Leopard" version
3. Install dmg2img
4. Run (as root): ./tools/mount_xcode_image.sh /path/to/xcode.dmg
5. Follow the instructions printed by ./tools/mount_xcode_image.sh
6. Copy or move the SDK into the tarballs/ directory

• Clang:

  • 32 bit: o32-clang++ test.cpp -O3 -o test OR i386-apple-darwinXX-clang++ test.cpp -O3 -o test
  • 64 bit: o64-clang++ test.cpp -O3 -o test OR x86_64-apple-darwinXX-clang++ test.cpp -O3 -o test

• GCC:

  • 32 bit: o32-g++ test.cpp -O3 -o test OR i386-apple-darwinXX-g++ test.cpp -O3 -o test
  • 64 bit: o64-g++ test.cpp -O3 -o test OR x86_64-apple-darwinXX-g++ test.cpp -O3 -o test

XX= the target version, you can find it out by running osxcross-conf and then see TARGET.

You can use the shortcuts o32-... for i386-apple-darwin..., depending on which you prefer.

I'll continue from here on with o32-clang, but remember, you can simply replace it with o32-gcc or i386-apple-darwin....

• make CC=o32-clang CXX=o32-clang++

• CC=o32-clang CXX=o32-clang++ ./configure --host=i386-apple-darwinXX

Note: libc++ requires Mac OS X 10.7 or newer! If you really need C++11 for an older OS X version, then you can do the following:

1. Build GCC so you have an up-to-date libstdc++
2. Build your source code with GCC or clang++-gstdc++ / clang++ -foc-use-gcc-libstdc++

Usage Examples:

• Clang:

  • C++98: o32-clang++ -stdlib=libc++ test.cpp -o test
  • C++11: o32-clang++ -stdlib=libc++ -std=c++11 test1.cpp -o test
  • C++14: o32-clang++ -stdlib=libc++ -std=c++14 test1.cpp -o test
  • C++1z: o32-clang++ -stdlib=libc++ -std=c++1z test1.cpp -o test

• Clang (shortcut):

  • C++98: o32-clang++-libc++ test.cpp -o test
  • C++11: o32-clang++-libc++ -std=c++11 test.cpp -o test
  • C++14: o32-clang++-libc++ -std=c++14 test.cpp -o test
  • C++1z: o32-clang++-libc++ -std=c++1z test.cpp -o test

• GCC

  • C++11: o32-g++-libc++ -std=c++11 test.cpp
  • C++14: o32-g++-libc++ -std=c++14 test.cpp -o test
  • C++1z: o32-g++-libc++ -std=c++1z test.cpp -o test

• build the first object file: o32-clang++ test1.cpp -O3 -flto -c
• build the second object file: o32-clang++ test2.cpp -O3 -flto -c
• link them with LTO: o32-clang++ -O3 -flto test1.o test2.o -o test

• Clang:
  • o64-clang++ test.cpp -O3 -arch i386 -arch x86_64 -o test
• GCC:
  • build the 32 bit binary: o32-g++ test.cpp -O3 -o test.i386
  • build the 64 bit binary: o64-g++ test.cpp -O3 -o test.x86_64
  • use lipo to generate the universal binary: x86_64-apple darwinXX-lipo -create test.i386 test.x86_64 -output test

The default deployment target is Mac OS X 10.5.

However, there are several ways to override the default value:

1. by passing OSX_VERSION_MIN=10.x to ./build.sh
2. by passing -mmacosx-version-min=10.x to the compiler
3. by setting the MACOSX_DEPLOYMENT_TARGET environment variable

>= 10.9 also defaults to libc++ instead of libstdc++, this behavior can be overriden by explicitly passing -stdlib=libstdc++ to clang.

x86_64h defaults to Mac OS X 10.8 and requires clang 3.5+. x86_64h = x86_64 with optimizations for the Intel Haswell Architecture.

You can build OSXCross with GCC this way:

CC=gcc CXX=g++ ./build.sh

You will need gcc/g++/gcc-objc 4.7+.

• multiarch/crossbuild: various cross-compilers (Systems: Linux, OS X, Windows, Archs: x86_64, i386, arm, ppc, mips) in Docker. OSXCross powers the Darwin builds.
• Smartmontools

• scripts/wrapper: GPLv2
• cctools/ld64: APSL 2.0
• xar: New BSD

• cjacker for the initial cctools port