Merge pull request #311777 from rhelmot/freebsd-minimal3/llvmPackages

llvmPackages: add support for native FreeBSD
2026-10-02 04:50:21 +00:00 · 2024-05-27 13:41:31 -04:00
parent 70f855d67d 3095f2de2e
commit 85536b3681
5 changed files with 98 additions and 42 deletions
--- a/pkgs/development/compilers/llvm/16/compiler-rt/asan-offset.patch
+++ b/pkgs/development/compilers/llvm/16/compiler-rt/asan-offset.patch
@@ -0,0 +1,11 @@
+--- a/lib/asan/CMakeLists.txt        2022-06-22 16:46:24.000000000 +0000
+++ b/lib/asan/CMakeLists.txt
+@@ -46,7 +46,7 @@ set(ASAN_STATIC_SOURCES
+   asan_rtl_static.cpp
+   )
+ 
+-if (NOT WIN32 AND NOT APPLE)
+if (LINUX)
+   list(APPEND ASAN_STATIC_SOURCES
+     asan_rtl_x86_64.S
+   )
--- a/pkgs/development/compilers/llvm/16/compiler-rt/freebsd-i386.patch
+++ b/pkgs/development/compilers/llvm/16/compiler-rt/freebsd-i386.patch
@@ -0,0 +1,21 @@
+--- a/lib/builtins/fp_lib.h	1969-12-31 16:00:01.000000000 -0800
+++ b/lib/builtins/fp_lib.h	2023-12-21 23:39:36.066927293 -0800
+@@ -26,18 +26,6 @@
+ #include <stdbool.h>
+ #include <stdint.h>
+ 
+-// x86_64 FreeBSD prior v9.3 define fixed-width types incorrectly in
+-// 32-bit mode.
+-#if defined(__FreeBSD__) && defined(__i386__)
+-#include <sys/param.h>
+-#if __FreeBSD_version < 903000 // v9.3
+-#define uint64_t unsigned long long
+-#define int64_t long long
+-#undef UINT64_C
+-#define UINT64_C(c) (c##ULL)
+-#endif
+-#endif
+-
+ #if defined SINGLE_PRECISION
+ 
+ typedef uint16_t half_rep_t;
--- a/pkgs/development/compilers/llvm/common/compiler-rt/default.nix
+++ b/pkgs/development/compilers/llvm/common/compiler-rt/default.nix
@@ -128,11 +128,17 @@ stdenv.mkDerivation ({
  '') + ''
    substituteInPlace lib/builtins/int_util.c \
      --replace "#include <stdlib.h>" ""
+  '' + (if stdenv.hostPlatform.isFreeBSD then
+    # As per above, but in FreeBSD assert is a macro and simply allowing it to be implicitly declared causes Issues!!!!!
+    ''
+    substituteInPlace lib/builtins/clear_cache.c lib/builtins/cpu_model.c \
+      --replace "#include <assert.h>" "#define assert(e) ((e)?(void)0:__assert(__FUNCTION__,__FILE__,__LINE__,#e))"
+    '' else ''
    substituteInPlace lib/builtins/clear_cache.c \
      --replace "#include <assert.h>" ""
    substituteInPlace lib/builtins/cpu_model${lib.optionalString (lib.versionAtLeast version "18") "/x86"}.c \
      --replace "#include <assert.h>" ""
-  '');
+  ''));

  # Hack around weird upsream RPATH bug
  postInstall = lib.optionalString (stdenv.hostPlatform.isDarwin) ''
--- a/pkgs/development/compilers/llvm/common/libcxx/default.nix
+++ b/pkgs/development/compilers/llvm/common/libcxx/default.nix
@@ -13,12 +13,16 @@
 , python3
 , fixDarwinDylibNames
 , version
-, cxxabi ? if stdenv.hostPlatform.isFreeBSD then libcxxrt else null
+, cxxabi ? null
 , libcxxrt
 , libunwind
 , enableShared ? !stdenv.hostPlatform.isStatic
 }:

+# note: our setup using libcxxabi instead of libcxxrt on FreeBSD diverges from
+# normal FreeBSD. This may cause issues with binary patching down the line.
+# If this becomes an issue, try adding as symlink libcxxrt.so -> libc++abi.so
+
 # external cxxabi is not supported on Darwin as the build will not link libcxx
 # properly and not re-export the cxxabi symbols into libcxx
 # https://github.com/NixOS/nixpkgs/issues/166205
--- a/pkgs/development/compilers/llvm/common/llvm/default.nix
+++ b/pkgs/development/compilers/llvm/common/llvm/default.nix
@@ -39,6 +39,7 @@
  # broken for the armv7l builder
  && !stdenv.hostPlatform.isAarch
 , enablePolly ? lib.versionAtLeast release_version "14"
+, enableTerminfo ? true
 }:

 let
@@ -124,42 +125,25 @@ stdenv.mkDerivation (rec {
    substituteInPlace cmake/modules/AddLLVM.cmake \
      --replace 'set(_install_name_dir INSTALL_NAME_DIR "@rpath")' "set(_install_name_dir)" \
      --replace 'set(_install_rpath "@loader_path/../''${CMAKE_INSTALL_LIBDIR}''${LLVM_LIBDIR_SUFFIX}" ''${extra_libdir})' ""
-  '' + (optionalString (lib.versionAtLeast release_version "15") (''
-
+  '' +
    # As of LLVM 15, marked as XFAIL on arm64 macOS but lit doesn't seem to pick
    # this up: https://github.com/llvm/llvm-project/blob/c344d97a125b18f8fed0a64aace73c49a870e079/llvm/test/MC/ELF/cfi-version.ll#L7
+    (optionalString (lib.versionAtLeast release_version "15") (''
    rm test/MC/ELF/cfi-version.ll

+  '' +
    # This test tries to call `sw_vers` by absolute path (`/usr/bin/sw_vers`)
    # and thus fails under the sandbox:
-  '' + (if lib.versionAtLeast release_version "16" then ''
+    (if lib.versionAtLeast release_version "16" then ''
    substituteInPlace unittests/TargetParser/Host.cpp \
      --replace '/usr/bin/sw_vers' "${(builtins.toString darwin.DarwinTools) + "/bin/sw_vers" }"
  '' else ''
    substituteInPlace unittests/Support/Host.cpp \
      --replace '/usr/bin/sw_vers' "${(builtins.toString darwin.DarwinTools) + "/bin/sw_vers" }"
-  '') + optionalString (lib.versionAtLeast release_version "16") ''
-
+  '') +
    # This test tries to call the intrinsics `@llvm.roundeven.f32` and
    # `@llvm.roundeven.f64` which seem to (incorrectly?) lower to `roundevenf`
-    # and `roundeven` on macOS.
-    #
-    # However these functions are glibc specific so the test fails:
-    #   - https://www.gnu.org/software/gnulib/manual/html_node/roundevenf.html
-    #   - https://www.gnu.org/software/gnulib/manual/html_node/roundeven.html
-    #
-    substituteInPlace test/ExecutionEngine/Interpreter/intrinsics.ll \
-      --replace "%roundeven32 = call float @llvm.roundeven.f32(float 0.000000e+00)" "" \
-      --replace "%roundeven64 = call double @llvm.roundeven.f64(double 0.000000e+00)" ""
-  '' + optionalString (!stdenv.hostPlatform.isx86 && lib.versionAtLeast release_version "18") ''
-
-    # fails when run in sandbox
-    substituteInPlace unittests/Support/VirtualFileSystemTest.cpp \
-      --replace "PhysicalFileSystemWorkingDirFailure" "DISABLED_PhysicalFileSystemWorkingDirFailure"
-  ''))) + optionalString (stdenv.isDarwin && stdenv.hostPlatform.isx86 && lib.versionAtLeast release_version "15") (optionalString (lib.versionOlder release_version "16") ''
-    # This test tries to call the intrinsics `@llvm.roundeven.f32` and
-    # `@llvm.roundeven.f64` which seem to (incorrectly?) lower to `roundevenf`
-    # and `roundeven` on x86_64 macOS.
+    # and `roundeven` on macOS and FreeBSD.
    #
    # However these functions are glibc specific so the test fails:
    #   - https://www.gnu.org/software/gnulib/manual/html_node/roundevenf.html
@@ -167,16 +151,28 @@ stdenv.mkDerivation (rec {
    #
    # TODO(@rrbutani): this seems to run fine on `aarch64-darwin`, why does it
    # pass there?
+    optionalString (lib.versionAtLeast release_version "16") ''
+    substituteInPlace test/ExecutionEngine/Interpreter/intrinsics.ll \
+      --replace "%roundeven32 = call float @llvm.roundeven.f32(float 0.000000e+00)" "" \
+      --replace "%roundeven64 = call double @llvm.roundeven.f64(double 0.000000e+00)" ""
+  '' +
+    # fails when run in sandbox
+    optionalString (!stdenv.hostPlatform.isx86 && lib.versionAtLeast release_version "18") ''
+    substituteInPlace unittests/Support/VirtualFileSystemTest.cpp \
+      --replace "PhysicalFileSystemWorkingDirFailure" "DISABLED_PhysicalFileSystemWorkingDirFailure"
+  ''))) +
+    # dup of above patch with different conditions
+    optionalString (stdenv.isDarwin && stdenv.hostPlatform.isx86 && lib.versionAtLeast release_version "15") (optionalString (lib.versionOlder release_version "16") ''
    substituteInPlace test/ExecutionEngine/Interpreter/intrinsics.ll \
      --replace "%roundeven32 = call float @llvm.roundeven.f32(float 0.000000e+00)" "" \
      --replace "%roundeven64 = call double @llvm.roundeven.f64(double 0.000000e+00)" ""

-  '' + ((optionalString (lib.versionAtLeast release_version "18") ''
-
+  '' +
    # fails when run in sandbox
+    ((optionalString (lib.versionAtLeast release_version "18") ''
    substituteInPlace unittests/Support/VirtualFileSystemTest.cpp \
      --replace "PhysicalFileSystemWorkingDirFailure" "DISABLED_PhysicalFileSystemWorkingDirFailure"
-  '') + ''
+  '') +
    # This test fails on darwin x86_64 because `sw_vers` reports a different
    # macOS version than what LLVM finds by reading
    # `/System/Library/CoreServices/SystemVersion.plist` (which is passed into
@@ -205,19 +201,20 @@ stdenv.mkDerivation (rec {
    # not clear to me when/where/for what this even gets used in LLVM.
    #
    # TODO(@rrbutani): fix/follow-up
-  '' + (if lib.versionAtLeast release_version "16" then ''
+    (if lib.versionAtLeast release_version "16" then ''
    substituteInPlace unittests/TargetParser/Host.cpp \
      --replace "getMacOSHostVersion" "DISABLED_getMacOSHostVersion"
  '' else ''
    substituteInPlace unittests/Support/Host.cpp \
      --replace "getMacOSHostVersion" "DISABLED_getMacOSHostVersion"
-  '') + ''
-
+  '') +
    # This test fails with a `dysmutil` crash; have not yet dug into what's
    # going on here (TODO(@rrbutani)).
+  ''
    rm test/tools/dsymutil/ARM/obfuscated.test
-  '')) + ''
+  '')) +
    # FileSystem permissions tests fail with various special bits
+  ''
    substituteInPlace unittests/Support/CMakeLists.txt \
      --replace "Path.cpp" ""
    rm unittests/Support/Path.cpp
@@ -229,22 +226,24 @@ stdenv.mkDerivation (rec {
  '' + lib.optionalString (lib.versionOlder release_version "13") ''
    # TODO: Fix failing tests:
    rm test/DebugInfo/X86/vla-multi.ll
-  '' + lib.optionalString (lib.versionAtLeast release_version "16") (''
-
+  '' +
    # Fails in the presence of anti-virus software or other intrusion-detection software that
    # modifies the atime when run. See #284056.
+    lib.optionalString (lib.versionAtLeast release_version "16") (''
    rm test/tools/llvm-objcopy/ELF/strip-preserve-atime.test
  '' + lib.optionalString (lib.versionOlder release_version "17") ''

-  '') + lib.optionalString (lib.versionAtLeast release_version "15" && lib.versionOlder release_version "17") ''
+  '') +
    # timing-based tests are trouble
+    lib.optionalString (lib.versionAtLeast release_version "15" && lib.versionOlder release_version "17") ''
    rm utils/lit/tests/googletest-timeout.py
-  '' + optionalString stdenv.hostPlatform.isMusl ''
+  '' +
+    # valgrind unhappy with musl or glibc, but fails w/musl only
+    optionalString stdenv.hostPlatform.isMusl ''
    patch -p1 -i ${./TLI-musl.patch}
    substituteInPlace unittests/Support/CMakeLists.txt \
      --replace "add_subdirectory(DynamicLibrary)" ""
    rm unittests/Support/DynamicLibrary/DynamicLibraryTest.cpp
-    # valgrind unhappy with musl or glibc, but fails w/musl only
    rm test/CodeGen/AArch64/wineh4.mir
  '' + optionalString stdenv.hostPlatform.isAarch32 ''
    # skip failing X86 test cases on 32-bit ARM
@@ -254,18 +253,28 @@ stdenv.mkDerivation (rec {
    rm test/tools/dsymutil/X86/op-convert.test
    rm test/tools/gold/X86/split-dwarf.ll
    rm test/tools/llvm-objcopy/MachO/universal-object.test
-  '' + optionalString (stdenv.hostPlatform.system == "armv6l-linux") ''
+  '' +
    # Seems to require certain floating point hardware (NEON?)
+    optionalString (stdenv.hostPlatform.system == "armv6l-linux") ''
    rm test/ExecutionEngine/frem.ll
+  '' +
+    # 1. TODO: Why does this test fail on FreeBSD?
+    # It seems to reference /usr/local/lib/libfile.a, which is clearly a problem.
+    # 2. This test fails for the same reason it fails on MacOS, but the fix is
+    # not trivial to apply.
+    optionalString stdenv.isFreeBSD ''
+    rm test/tools/llvm-libtool-darwin/L-and-l.test
+    rm test/ExecutionEngine/Interpreter/intrinsics.ll
  '' + ''
    patchShebangs test/BugPoint/compile-custom.ll.py
-  '' + (lib.optionalString (lib.versionOlder release_version "13") ''
+  '' +
    # Tweak tests to ignore namespace part of type to support
    # gcc-12: https://gcc.gnu.org/PR103598.
    # The change below mangles strings like:
    #    CHECK-NEXT: Starting llvm::Function pass manager run.
    # to:
    #    CHECK-NEXT: Starting {{.*}}Function pass manager run.
+    (lib.optionalString (lib.versionOlder release_version "13") (''
    for f in \
      test/Other/new-pass-manager.ll \
      test/Other/new-pm-O0-defaults.ll \
@@ -286,18 +295,19 @@ stdenv.mkDerivation (rec {
        --replace 'Starting llvm::' 'Starting {{.*}}' \
        --replace 'Finished llvm::' 'Finished {{.*}}'
    done
+  '' +
    # gcc-13 fix
+  ''
    sed -i '/#include <string>/i#include <cstdint>' \
      include/llvm/DebugInfo/Symbolize/DIPrinter.h
-  '');
+  ''));

+  # Workaround for configure flags that need to have spaces
  preConfigure = if lib.versionAtLeast release_version "15" then ''
-    # Workaround for configure flags that need to have spaces
    cmakeFlagsArray+=(
      -DLLVM_LIT_ARGS="-svj''${NIX_BUILD_CORES} --no-progress-bar"
    )
  '' else ''
-    # Workaround for configure flags that need to have spaces
    cmakeFlagsArray+=(
      -DLLVM_LIT_ARGS='-svj''${NIX_BUILD_CORES} --no-progress-bar'
    )
@@ -332,9 +342,13 @@ stdenv.mkDerivation (rec {
    "-DLLVM_HOST_TRIPLE=${stdenv.hostPlatform.config}"
    "-DLLVM_DEFAULT_TARGET_TRIPLE=${stdenv.hostPlatform.config}"
    "-DLLVM_ENABLE_DUMP=ON"
+    (lib.cmakeBool "LLVM_ENABLE_TERMINFO" enableTerminfo)
+  ] ++ optionals (!doCheck) [
+    "-DLLVM_INCLUDE_TESTS=OFF"
  ] ++ optionals stdenv.hostPlatform.isStatic [
    # Disables building of shared libs, -fPIC is still injected by cc-wrapper
    "-DLLVM_ENABLE_PIC=OFF"
+    "-DCMAKE_SKIP_INSTALL_RPATH=ON"
    "-DLLVM_BUILD_STATIC=ON"
    # libxml2 needs to be disabled because the LLVM build system ignores its .la
    # file and doesn't link zlib as well.