Building LLVM and Clang on macOS 10.14.6 with Xcode 11.1
I recently attended the 2019 LLVM Developer Meeting:
Sorry, X (nÊe Twitter) deleted my post.
After the first hour, spent listening to the keynote called “Generating Optimized Code with GlobalISel” my brain hurt, but… it was that good kind of pain, like the one you get after a particularly hard workout. By the end of the day, after having seen several more talks, I was psyched up!
Naturally, when I went home and immediately tried to build LLVM on my macBook Pro running macOS 10.14.6 Mojave, and with Xcode 11.1 installed.
Before We Begin
Make sure you’ve launched Xcode at least once and it’s completed the First Launch Experience, which includes accepting the license agreement. You can also accept using the command line:
$ sudo xcodebuild -license accept
System Headers
I’m runing Xcode 11.1, which comes with the macOS 10.15 Catalina SDK. According to the Xcode 10.0 release notes, Apple has helpfully provided a package to install the system headers into /usr/include. You’ll need these for the LLVM unit tests.
$ open /Library/Developer/CommandLineTools/Packages/macOS_SDK_headers_for_macOS_10.14.pkg
This command will open the package installer UI for you so you can install the headers:
Install Requirements
Homebrew has some flaws, but I just want dependency management to get out of my way and let me do work, and it’s really excellent at that.
I updated Homebrew and upgraded all of my installed software. What can I say? I like to live on the edge.
$ brew update
[...]
$ brew upgrade
==> Upgrading 9 outdated packages:
[...]
$
This is all I had to do to get the tools I needed to build LLVM:
$ brew install cmake
==> Downloading https://homebrew.bintray.com/bottles/cmake-3.15.4.mojave.bottle.tar.gz
Already downloaded: /Users/liamr/Library/Caches/Homebrew/downloads/9013e9f40f45899dbcfa580fa058dfd988d4f12a2d900b9d8b79c4ea99cba1de--cmake-3.15.4.mojave.bottle.tar.gz
==> Pouring cmake-3.15.4.mojave.bottle.tar.gz
==> Caveats
Emacs Lisp files have been installed to:
/usr/local/share/emacs/site-lisp/cmake
==> Summary
đē /usr/local/Cellar/cmake/3.15.4: 5,800 files, 53.4MB
$ brew install ninja
[...]
$ brew install ccache
[...]
$
Clone, Configure, Build & Test
Next, it was time to clone the repository.
$ git clone git@github.com:llvm/llvm-project.git
Cloning into 'llvm-project'...
remote: Enumerating objects: 1141, done.
remote: Counting objects: 100% (1141/1141), done.
remote: Compressing objects: 100% (155/155), done.
remote: Total 3577397 (delta 998), reused 1011 (delta 981), pack-reused 3576256
Receiving objects: 100% (3577397/3577397), 1.14 GiB | 5.98 MiB/s, done.
Resolving deltas: 100% (2862094/2862094), done.
Checking out files: 100% (86709/86709), done.
$
Whoo-baby, that’s a monorepo behemoth. Next, I configured a build.
$ cd llvm-project
$ mkdir build
$ cd build/
$ cmake -G Ninja ../llvm/ \
-DLLVM_ENABLE_PROJECTS="clang" \
-DCMAKE_BUILD_TYPE="Release" \
-DLLVM_ENABLE_ASSERTIONS=On \
-DLLVM_CCACHE_BUILD=On
[...]
-- Build files have been written to: /Users/liamr/Code/llvm-project/build
$
And now, we build! (And build and run all the tests.)
$ ninja check-all
[184/4275] Generating VCSRevision.h
-- Found Git: /usr/bin/git (found version "2.21.0 (Apple Git-122)")
[2412/4275] Generating VCSVersion.inc
-- Found Git: /usr/bin/git (found version "2.21.0 (Apple Git-122)")
[3470/4275] cd /Users/liamr/Code/llvm-project/clang/bindings/python && /usr/local/Cellar/cmake/3.15.4/...E env CLANG_LIBRARY_PATH=/Users/liamr/Code/llvm-project/build/lib /usr/bin/python -m unittest discover
.................................................s................................................s.s.......s...s.........s...
----------------------------------------------------------------------
Ran 126 tests in 0.693s
OK (skipped=6)
[4274/4275] Running all regression tests
llvm-lit: /Users/liamr/Code/llvm-project/llvm/utils/lit/lit/llvm/config.py:342: note: using clang: /Users/liamr/Code/llvm-project/build/bin/clang
llvm-lit: /Users/liamr/Code/llvm-project/llvm/utils/lit/lit/util.py:408: note: using SDKROOT: u'/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.15.sdk'
llvm-lit: /Users/liamr/Code/llvm-project/build/utils/lit/tests/lit.cfg:69: warning: Setting a timeout per test not supported. Requires the Python psutil module but it could not be found. Try installing it via pip or via your operating system's package manager. Some tests will be skipped and the --timeout command line argument will not work.
Testing Time: 680.16s
Expected Passes : 48896
Expected Failures : 172
Unsupported Tests : 1191
1 warning(s) in tests.
$
Test Compilation
About an hour later, my laptop’s fans have had a great workout, and I’ve built something! But… does it work?
$ mkdir outputs
$ mkdir inputs
$ vi inputs/helloworld.c
[...]
$ bat inputs/helloworld.c
ââââââââŦâââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââ
â File: inputs/helloworld.c
ââââââââŧâââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââ
1 â #include <stdio.h>
2 â
3 â int main() {
4 â printf("Hello, World!\n");
5 â return 0;
6 â }
ââââââââ´âââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââ
$ bin/clang -Wall inputs/helloworld.c -o outputs/helloworld
$ outputs/helloworld
Hello, World!
$ bin/clang -S -emit-llvm -O0 inputs/helloworld.c -o outputs/helloworld.ll
$ bat outputs/helloworld.ll
ââââââââŦâââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââ
â File: outputs/helloworld.ll
ââââââââŧâââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââ
1 â ; ModuleID = 'inputs/helloworld.c'
2 â source_filename = "inputs/helloworld.c"
3 â target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
4 â target triple = "x86_64-apple-macosx10.14.6"
5 â
6 â @.str = private unnamed_addr constant [15 x i8] c"Hello, World!\0A\00", align 1
7 â
8 â ; Function Attrs: noinline nounwind optnone ssp uwtable
9 â define i32 @main() #0 {
10 â entry:
11 â %retval = alloca i32, align 4
12 â store i32 0, i32* %retval, align 4
13 â %call = call i32 (i8*, ...) @printf(i8* getelementptr inbounds ([15 x i8], [15 x i8]* @.str, i64 0, i64 0))
14 â ret i32 0
15 â }
16 â
17 â declare i32 @printf(i8*, ...) #1
18 â
19 â attributes #0 = { noinline nounwind optnone ssp uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="fals
â e" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-prot
â ector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+cx8,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
20 â attributes #1 = { "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="all" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-
â fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+cx8,+fxsr,+mmx,+sahf,
â +sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
21 â
22 â !llvm.module.flags = !{!0, !1, !2}
23 â !llvm.ident = !{!3}
24 â
25 â !0 = !{i32 2, !"SDK Version", [2 x i32] [i32 10, i32 15]}
26 â !1 = !{i32 1, !"wchar_size", i32 4}
27 â !2 = !{i32 7, !"PIC Level", i32 2}
28 â !3 = !{!"clang version 10.0.0 (git@github.com:llvm/llvm-project.git 23fdd513a3ba7594ccdab1ea30608f4cab7faee4)"}
ââââââââ´âââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââââ
$
Success
IT LIVES!
YMMV, but this worked for me. Let me know on Twitter if you run into anything amiss!