haskell.compiler.ghc*: use matching CLANG & CC on Darwin with LLVM

The Darwin LLVM backend of GHC (which is mostly interesting for GHC < 9.2) uses clang as configured via the CLANG environment variable as an assembler. Since it processes outputs of clang as configured via the CC variable, we need to make sure these versions match or risk CLANG clang not understanding the output of CC clang. In the past this wasn't really a problem as due to the fairly old default clang version in the stdenv, clang 11 would be used for CC. CLANG would always be a newer version and deal with the output without any problems. Ever since the upgrade of the default clang version for darwin (bcbdb80), CC would often be a newer version of clang than CLANG, causing build problems in some packages like crypton (for GHC 8.10.7 and 9.0.2 on aarch64-darwin where the darwin LLVM backend was actually used).
NixOS · Sep 8, 2024 · 19c88fb · 19c88fb
1 parent 2dae227
commit 19c88fb
Show file tree

Hide file tree

Showing 3 changed files with 60 additions and 9 deletions.
diff --git a/pkgs/development/compilers/ghc/8.10.7.nix b/pkgs/development/compilers/ghc/8.10.7.nix
@@ -184,6 +184,9 @@ let
           if stdenv.targetPlatform.isDarwin
           then cc.bintools
           else cc.bintools.bintools;
+
+        # clang is used as an assembler on darwin with the LLVM backend
+        clang = cc;
       }.${name};
     in
     "${tools}/bin/${tools.targetPrefix}${name}";
@@ -335,7 +338,17 @@ stdenv.mkDerivation (rec {
     export OPT="${lib.getBin buildTargetLlvmPackages.llvm}/bin/opt"
   '' + lib.optionalString (useLLVM && stdenv.targetPlatform.isDarwin) ''
     # LLVM backend on Darwin needs clang: https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/codegens.html#llvm-code-generator-fllvm
-    export CLANG="${buildTargetLlvmPackages.clang}/bin/${buildTargetLlvmPackages.clang.targetPrefix}clang"
+    # The executable we specify via $CLANG is used as an assembler (exclusively, it seems, but this isn't
+    # clarified in any user facing documentation). As such, it'll be called on assembly produced by $CC
+    # which usually comes from the darwin stdenv. To prevent a situation where $CLANG doesn't understand
+    # the assembly it is given, we need to make sure that it matches the LLVM version of $CC if possible.
+    # It is unclear (at the time of writing 2024-09-01)  whether $CC should match the LLVM version we use
+    # for llc and opt which would require using a custom darwin stdenv for targetCC.
+    export CLANG="${
+      if targetCC.isClang
+      then toolPath "clang" targetCC
+      else "${buildTargetLlvmPackages.clang}/bin/${buildTargetLlvmPackages.clang.targetPrefix}clang"
+    }"
   '' + ''
     # No need for absolute paths since these tools only need to work during the build
     export CC_STAGE0="$CC_FOR_BUILD"
@@ -477,10 +490,14 @@ stdenv.mkDerivation (rec {
       "LLVM llc command" "${lib.getBin llvmPackages.llvm}/bin/llc" \
       "LLVM opt command" "${lib.getBin llvmPackages.llvm}/bin/opt"
   ''
-  # FIXME(@sternenseemann): use installCC instead if possible
   + lib.optionalString (useLLVM && stdenv.targetPlatform.isDarwin) ''
     ghc-settings-edit "$settingsFile" \
-      "LLVM clang command" "${llvmPackages.clang}/bin/${llvmPackages.clang.targetPrefix}clang"
+      "LLVM clang command" "${
+        # See comment for CLANG in preConfigure
+        if installCC.isClang
+        then toolPath "clang" installCC
+        else "${llvmPackages.clang}/bin/${llvmPackages.clang.targetPrefix}clang"
+      }"
   ''
   + ''
 

diff --git a/pkgs/development/compilers/ghc/common-hadrian.nix b/pkgs/development/compilers/ghc/common-hadrian.nix
@@ -320,6 +320,9 @@ let
           if stdenv.targetPlatform.isDarwin
           then cc.bintools
           else cc.bintools.bintools;
+
+        # clang is used as an assembler on darwin with the LLVM backend
+        clang = cc;
       }.${name};
     in
     "${tools}/bin/${tools.targetPrefix}${name}";
@@ -409,7 +412,17 @@ stdenv.mkDerivation ({
     export OPT="${lib.getBin buildTargetLlvmPackages.llvm}/bin/opt"
   '' + lib.optionalString (useLLVM && stdenv.targetPlatform.isDarwin) ''
     # LLVM backend on Darwin needs clang: https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/codegens.html#llvm-code-generator-fllvm
-    export CLANG="${buildTargetLlvmPackages.clang}/bin/${buildTargetLlvmPackages.clang.targetPrefix}clang"
+    # The executable we specify via $CLANG is used as an assembler (exclusively, it seems, but this isn't
+    # clarified in any user facing documentation). As such, it'll be called on assembly produced by $CC
+    # which usually comes from the darwin stdenv. To prevent a situation where $CLANG doesn't understand
+    # the assembly it is given, we need to make sure that it matches the LLVM version of $CC if possible.
+    # It is unclear (at the time of writing 2024-09-01)  whether $CC should match the LLVM version we use
+    # for llc and opt which would require using a custom darwin stdenv for targetCC.
+    export CLANG="${
+      if targetCC.isClang
+      then toolPath "clang" targetCC
+      else "${buildTargetLlvmPackages.clang}/bin/${buildTargetLlvmPackages.clang.targetPrefix}clang"
+    }"
   '' +
   lib.optionalString (stdenv.isLinux && hostPlatform.libc == "glibc") ''
     export LOCALE_ARCHIVE="${glibcLocales}/lib/locale/locale-archive"
@@ -627,10 +640,14 @@ stdenv.mkDerivation ({
       "LLVM llc command" "${lib.getBin llvmPackages.llvm}/bin/llc" \
       "LLVM opt command" "${lib.getBin llvmPackages.llvm}/bin/opt"
   ''
-  # FIXME(@sternenseemann): use installCC instead if possible
   + lib.optionalString (useLLVM && stdenv.targetPlatform.isDarwin) ''
     ghc-settings-edit "$settingsFile" \
-      "LLVM clang command" "${llvmPackages.clang}/bin/${llvmPackages.clang.targetPrefix}clang"
+      "LLVM clang command" "${
+        # See comment for CLANG in preConfigure
+        if installCC.isClang
+        then toolPath "clang" installCC
+        else "${llvmPackages.clang}/bin/${llvmPackages.clang.targetPrefix}clang"
+      }"
   ''
   # Work around a GHC bug which causes unlit to be installed under a different
   # name than is used in the settings file.

diff --git a/pkgs/development/compilers/ghc/common-make-native-bignum.nix b/pkgs/development/compilers/ghc/common-make-native-bignum.nix
@@ -187,6 +187,9 @@ let
           if stdenv.targetPlatform.isDarwin
           then cc.bintools
           else cc.bintools.bintools;
+
+        # clang is used as an assembler on darwin with the LLVM backend
+        clang = cc;
       }.${name};
     in
     "${tools}/bin/${tools.targetPrefix}${name}";
@@ -335,7 +338,17 @@ stdenv.mkDerivation (rec {
     export OPT="${lib.getBin buildTargetLlvmPackages.llvm}/bin/opt"
   '' + lib.optionalString (useLLVM && stdenv.targetPlatform.isDarwin) ''
     # LLVM backend on Darwin needs clang: https://downloads.haskell.org/~ghc/latest/docs/html/users_guide/codegens.html#llvm-code-generator-fllvm
-    export CLANG="${buildTargetLlvmPackages.clang}/bin/${buildTargetLlvmPackages.clang.targetPrefix}clang"
+    # The executable we specify via $CLANG is used as an assembler (exclusively, it seems, but this isn't
+    # clarified in any user facing documentation). As such, it'll be called on assembly produced by $CC
+    # which usually comes from the darwin stdenv. To prevent a situation where $CLANG doesn't understand
+    # the assembly it is given, we need to make sure that it matches the LLVM version of $CC if possible.
+    # It is unclear (at the time of writing 2024-09-01)  whether $CC should match the LLVM version we use
+    # for llc and opt which would require using a custom darwin stdenv for targetCC.
+    export CLANG="${
+      if targetCC.isClang
+      then toolPath "clang" targetCC
+      else "${buildTargetLlvmPackages.clang}/bin/${buildTargetLlvmPackages.clang.targetPrefix}clang"
+    }"
   ''
   + ''
     # No need for absolute paths since these tools only need to work during the build
@@ -494,10 +507,14 @@ stdenv.mkDerivation (rec {
       "LLVM llc command" "${lib.getBin llvmPackages.llvm}/bin/llc" \
       "LLVM opt command" "${lib.getBin llvmPackages.llvm}/bin/opt"
   ''
-  # FIXME(@sternenseemann): use installCC instead if possible
   + lib.optionalString (useLLVM && stdenv.targetPlatform.isDarwin) ''
     ghc-settings-edit "$settingsFile" \
-      "LLVM clang command" "${llvmPackages.clang}/bin/${llvmPackages.clang.targetPrefix}clang"
+      "LLVM clang command" "${
+        # See comment for CLANG in preConfigure
+        if installCC.isClang
+        then toolPath "clang" installCC
+        else "${llvmPackages.clang}/bin/${llvmPackages.clang.targetPrefix}clang"
+      }"
   ''
   + ''