Is there any way to compile x86-64 + arm64 universal binaries in macOS 10.14? - macos

The official way to compile universal binaries that target Apple's new M1 processor is to use Xcode 12, which requires at minimum macOS 10.15.4. I'm currently running macOS 10.14.6 though and do not want to upgrade if I can avoid it. Is there an unofficial way to compile such a universal binary when running macOS 10.14?

This is possible to do. Of course it can't be done with Xcode on its own since Apple has long since dropped support for macOS 10.14 in its developer tools, but can be down either by compiling from the command line or using another build system like cmake or qmake.
It at minimum requires two things:
A newer macOS SDK than is officially supported in macOS 10.14, obtained from Xcode 12.2 or later
A newer version of LLVM / clang that is compatible the macOS SDK from the version of Xcode you obtained
You can actually get both things by downloading the latest release of Xcode, since as of the time of this writing (December 2022) the release of LLVM included in the latest release of Xcode (14.2) will still run in macOS 10.14.
Should that ever stop working, though, or if you would prefer not to use Xcode's release of LLVM, you can also install a newer version of LLVM and clang using MacPorts, as MacPorts continues to support older macOS releases (unlike Homebrew). The command would be sudo port install clang-15, or whatever the latest available major release of clang is.
Another option is to built and install your own copy of LLVM and clang, and the steps to do so are laid out in LLVM's documentation.
For best results, the version of LLVM should be the same major version or newer than the version of LLVM / clang included in the release of Xcode you obtained.
Once you have both of those, you can compile C, C++, or Objective-C and probably Swift with a command like:
/path/to/newer/clang \
-arch x86_64 -arch arm64 \
-isysroot /path/to/Xcode14.2.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk \
source.m source2.cpp source3.c ...
If you installed clang using MacPorts, then its path would be something like /opt/local/bin/clang-mp-15.
In the case of only using a newer release of Xcode, you can do it like this:
/path/to/Xcode14.2.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang \
-arch x86_64 -arch arm64 \
-isysroot /path/to/Xcode14.2.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk \
source.m source2.cpp source3.c ...
I've only tested this with C, C++ and Objective-C so far, and not Swift, but presumably it will work for Swift as well.
Note that if you want to target macOS 10.14, it's still necessary to avoid using parts of the macOS SDK as well as C++ and Swift features that require macOS 10.15 or later.
(It may be possible to at least make use of newer C++ features in macOS 10.14 and earlier if you build your own universal binaries of LLVM's runtimes (namely libc++ and libc++abi) from the later release of LLVM, and then compile your code with the arguments -nostdlib, statically link to libc++ and libc++abi, and then explicitly link to /usr/lib/libSystem.B.dylib. But so far I haven't managed to get an arm64 copy of LLVM's runtimes to compile in macOS 10.14.)

Short answer: No
Why?
Apple did it. There's a doc from apple here
You will have to use Xcode 12.2 and above. You must upgrade to MacOS 10.15.4+ or 11 to benifit the latest feature.
Apple brings the tool lipo to convert binaires for ARM64 and AMD64 architectures.
With Apple silicon, you can compile binaries for ARM64 and AMD64. You can debug ARM64 binaries naturally. You can debug AMD64 binaries under Rosetta translation.
With Intel x86 chips, you can compile binaries for ARM64 and AMD64. You can debug AMD64 binaries naturally. But you cannot debug ARM64 binaries under Rosetta translation.
This means Apple Silicon's ARM64 architecture can do more on MacOS. It's unfair for AMD64 chips. Tipical Apple. Hopefully someone can come out and unseals the secret Apple silicon. But I don't think apple would like to share. If Intel/AMD wake up and bring us more powerful chips which can compile/debug ARM64 binaries for MacOS, apple wouldn't be happy with their sales. For the longterm, we won't be able to do it (seemingly). Apple kicked Nvidia away years ago, now Intel, soon AMD. Mark it here. 10 years later, let's see.

Related

Assembly armv8 on mac os

I would like to assemble Aarch64 armv8 Assembly on my mac and eventually run it with qemu and maybe on a real device like a raspberry pi 2 or 4 later on. I don't know how to assemble the code I'm going to write, gcc, llvm-gcc and clang don't seem to support the -arch=armv8 flag or anything similar. So I can't build for the targeted architecture, how could I achieve this?
I'm running mac os 10.14.5. I wouldn't mind finding a solution that works on a recent ubuntu version either since I have a VM for linux development.
The clang version that ships with Xcode supports -arch arm64. (Or armv7 for 32bit.)
Note that if you want to use any libraries though, they'll have to be arm64 as well. If you want, you can invoke the iOS toolchain with xcrun -sdk iphoneos clang -arch arm64 [...], but then you'll also have to pull the libraries you want off of some IPSW and stuff them into qemu.
Also note that the above will give you a Mach-O binary. For your Raspberry Pi, you'll probably want an ELF, and you'll probably want gcc rather than clang. You should be able to build both gcc and GNU binutils from source with --target as either aarch64-linux-gnu or aarch64-none-elf, depending on your goals. Yet another note: since macOS silently aliases gcc to clang and many tools depend on that, you'll probably also want to build this toolchain with something like --program-prefix=aarch64-.

Using GCC in -m32 mode on mac OS

I need to use gcc in -m32 mode on my macOS Mojave. I know it was deprecated some time ago. What should I do to make it possible?
You would need to download an older version of Xcode (eg. 9.x), and pre 10.14 SDK's (eg. 10.13), then try from there. Apple dropped Xcode capability/support for i386 in 10.14. In other words, Mojave can run existing 32-bit applications but can't build 32-bit applications anymore.
↳ MacPorts Discussion : User information about macOS Mojave
Works for me by installing:
/Library/Developer/CommandLineTools/Packages/macOS_SDK_headers_for_macOS_10.14.pkg
as commented here: Can't compile C program on a Mac after upgrade to Mojave

using command-line clang, OS X 10.9.1, Xcode 5.0.2

I'd like to use the -fsanitize=address features of clang.
I'm on:
OS X 10.9.1 with Xcode 5.0.2
Apple LLVM version 5.0 (clang-500.2.79) (based on LLVM 3.3svn)
I'm running clang from the command line, and I'm getting the error:
clang: error: unsupported argument 'address' to option 'fsanitize='
I'm told this works, but searching around hasn't shown me the way; do I need to bring in a different version of clang?
Thanks!
The out-of-box version of clang on OS X does not implement support of Address Sanitizer. You will have to build your own version of clang from sources.
You need a different build of clang than the one apple gave you. The simplest way to get one (or more) is to install Homebrew http://brew.sh
and then
brew install llvm --with-asan
You could also try a prebuilt clang from http://llvm.org/releases/download.htm.
It's by no means impossible to build clang 'by hand', but last I checked it did take multiple checkouts to get the source tree. The other options should have you asaning in minutes.

How can I update clang to 3.3 on Mac OS X 10.6

I'm running Mac OS X 10.6 and want to install TextMate 2, which is for Mac OS X 10.7+.
But all what it needs is a newer version of clang (LLVM), which is included in Lion and Mountain Lion. I read before here I can do that with MacPorts. So I used port install clang and MacPorts updated to clang-2.9, then clang-3.2 and finally clang-3.3. I thought, its updated now and I checked the version: clang --version. And it's not updated:
Apple clang version 2.0 (tags/Apple/clang-139) (based on LLVM 2.9svn)
Target: x86_64-apple-darwin10
Thread model: posix
So I copied the new clang file in this path (/opt/local/libexec/llvm-3.3/bin) to/usr/bin`. But now there's a fail :( :
dyld: Library not loaded: #executable_path/../lib/libLLVM-3.3svn.dylib
Referenced from: /usr/bin/clang
Reason: image not found Trace/BPT trap
The libLLVM-3.3svn.dylib is at /opt/local/libexec/llvm-3.3/lib. What can I do now, that
it runs clang-3.3? Sorry for my not perfect english ;) Thank you!
If you run clang from /usr/bin, it will be looking for the LLVM library in /usr/bin/../lib/libLLVM-3.3svn.dylib, i.e., /usr/lib/libLLVM-3.3svn.dylib (at least that's what the dynamic loader is telling you). You could try to copy the lib file into /usr/lib.
Alternatively, you can just download a more recent official build of Clang here:
http://llvm.org/releases/download.html
As far as I know, these builds are self-contained and do not need any dynamic libraries. However, it could be necessary to copy Apple's ARC libraries into a certain directory. If you run into trouble, please ask again.
Another approach is to build Clang and LLVM from source. You can find the repository URLs and instructions here:
http://clang.llvm.org/get_started.html
This should also build Clang with the LLVM libraries statically linked.
3.3, by the way, has not been officially released, so I would recommend using 3.2 unless you need any specific new features.

Obtaining GCC for OSX with Developer Tools installed

I want to start working with C++0x. I see that GCC 4.7 has a fair amount of functionality available. I already have XCode 3.2 installed in /Developer
I downloaded: http://fileboar.com/gcc/snapshots/LATEST-4.7/gcc-4.7-20110528.tar.bz2
Can I somehow compile this in /opt/gcc-4.7? How do I then work with my path so I can compile with GCC 4.7 from the command-line but have OSX use the version it needs?
OSX does not need gcc to run - the Developer tools are optional. So you only need to choose between gcc's when you compile. In Xcode you chose explicitly the gcc andin Makefiles you can set $(CC) or similar to the full path.
Alternatively rename the gcc-4.7 gcc to gcc-4.7 and use that so gcc is always the Apple one.
For ease of using multiple C++ compilers I use macports (or fink or homebrew) which will compile the compilers with the correct patches and also has a port select command to switch between the C++ compilers

Resources