feat: add clang support to llvm-core Conan recipe

conanfile.py

@@ -185,7 +185,7 @@ def requirements(self):
         )
         self.requires("arrow/15.0.1-oss", transitive_headers=True, transitive_libs=True)
         if self.options.get_safe("enable_jit"):
-            self.requires("llvm-core/19.1.7")
+            self.requires("llvm-core/19.1.7-bolt")
 
         if self.options.get_safe("enable_s3"):
             self.requires(
@@ -352,6 +352,7 @@ def configure(self):
             self.options[llvm_core].with_z3 = False
             self.options[llvm_core].with_zstd = False
             self.options[llvm_core].with_ffi = False
+            self.options[llvm_core].with_clang = True
 
         if self.options.get_safe("enable_hdfs") and self.options.get_safe(
             "use_arrow_hdfs"
@@ -451,6 +452,15 @@ def generate(self):
             tc.cache_variables["ENABLE_BOLT_JIT"] = "ON"
             tc.preprocessor_definitions["ENABLE_BOLT_JIT"] = 1
 
+            # Verify clang exists in llvm-core package.
+            llvm_dep = self.dependencies["llvm-core"]
+            clang_path = os.path.join(str(llvm_dep.package_folder), "bin", "clang")
+            if not os.path.exists(clang_path):
+                raise Exception(
+                    f"clang not found at {clang_path}. "
+                    "Ensure llvm-core is built with -o llvm-core/*:with_clang=True"
+                )
+
             # TODO: Refactor the IR codegen of expression evaluation
             # Disable it right now
             tc.cache_variables["ENABLE_BOLT_EXPR_JIT"] = "OFF"

scripts/conan/patches/llvm-core-add-clang-support.patch

@@ -0,0 +1,90 @@
+diff --git a/recipes/llvm-core/config.yml b/recipes/llvm-core/config.yml
+index ba4acd33..b1f0b2a1 100644
+--- a/recipes/llvm-core/config.yml
++++ b/recipes/llvm-core/config.yml
+@@ -1,5 +1,5 @@
+ versions:
+-  "19.1.7":
++  "19.1.7-bolt":
+     folder: all
+   "13.0.0":
+     folder: all
+diff --git a/recipes/llvm-core/all/conandata.yml b/recipes/llvm-core/all/conandata.yml
+index dd26cdba..bc4c16a9 100644
+--- a/recipes/llvm-core/all/conandata.yml
++++ b/recipes/llvm-core/all/conandata.yml
+@@ -1,11 +1,14 @@
+ sources:
+-  "19.1.7":
++  "19.1.7-bolt":
+     "llvm":
+       url: https://github.com/llvm/llvm-project/releases/download/llvmorg-19.1.7/llvm-19.1.7.src.tar.xz
+       sha256: 96f833c6ad99a3e8e1d9aca5f439b8fd2c7efdcf83b664e0af1c0712c5315910
+     "cmake":
+       url: https://github.com/llvm/llvm-project/releases/download/llvmorg-19.1.7/cmake-19.1.7.src.tar.xz
+       sha256: 11c5a28f90053b0c43d0dec3d0ad579347fc277199c005206b963c19aae514e3
++    "clang":
++      url: https://github.com/llvm/llvm-project/releases/download/llvmorg-19.1.7/clang-19.1.7.src.tar.xz
++      sha256: 11e5e4ecab5338b9914de3b83a4622cb200de466b7c56ba675afb72fa7d64675
+   "13.0.0":
+     url: https://github.com/llvm/llvm-project/releases/download/llvmorg-13.0.0/llvm-13.0.0.src.tar.xz
+     sha256: 408d11708643ea826f519ff79761fcdfc12d641a2510229eec459e72f8163020
+@@ -17,7 +20,7 @@ sources:
+     sha256: ce8508e318a01a63d4e8b3090ab2ded3c598a50258cc49e2625b9120d4c03ea5
+
+ patches:
+-  "19.1.7":
++  "19.1.7-bolt":
+     - patch_file: patches/19x/0000-cmake-dependencies.patch
+       patch_description: fix references to third party libs to match conan variables and targets
+       patch_type: conan
+diff --git a/recipes/llvm-core/all/conanfile.py b/recipes/llvm-core/all/conanfile.py
+index e8290099..fb18c9dd 100644
+--- a/recipes/llvm-core/all/conanfile.py
++++ b/recipes/llvm-core/all/conanfile.py
+@@ -171,6 +171,7 @@ class LLVMCoreConan(ConanFile):
+         "with_xml2": [True, False],
+         "with_z3": [True, False],
+         "with_zstd": [True, False],
++        "with_clang": [True, False],
+     }
+     default_options = {
+         "shared": False,
+@@ -193,6 +194,7 @@ class LLVMCoreConan(ConanFile):
+         "with_z3": True,
+         "with_zlib": True,
+         "with_zstd": True,
++        "with_clang": False,
+     }
+
+     @property
+@@ -301,6 +303,8 @@ class LLVMCoreConan(ConanFile):
+             # LLVM >=15 split up several components in its release, including cmake
+             get(self, **sources["llvm"], destination='llvm-main', strip_root=True)
+             get(self, **sources["cmake"], destination='cmake', strip_root=True)
++            if "clang" in sources:
++                get(self, **sources["clang"], destination='clang', strip_root=True)
+
+     def _apply_resource_limits(self, cmake_definitions):
+         if os.getenv("CONAN_CENTER_BUILD_SERVICE"):
+@@ -373,6 +377,9 @@ class LLVMCoreConan(ConanFile):
+         if self.options.targets != "all":
+             cmake_variables["LLVM_TARGETS_TO_BUILD"] = self.options.targets
+
++        if self.options.with_clang:
++            cmake_variables["LLVM_ENABLE_PROJECTS"] = "clang"
++
+         self._apply_resource_limits(cmake_variables)
+
+         if is_msvc(self):
+@@ -415,7 +422,9 @@ class LLVMCoreConan(ConanFile):
+             "LLVMTableGenGlobalISel.*",
+             "CONAN_LIB.*",
+             "LLVMExegesis.*",
+-            "LLVMCFIVerify.*"
++            "LLVMCFIVerify.*",
++            "clang.*",
++            "libclang.*",
+         ]
+         graphviz_options = textwrap.dedent(f"""
+             set(GRAPHVIZ_EXECUTABLES OFF)

scripts/install-bolt-deps.sh

@@ -19,7 +19,7 @@ CUR_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" > /dev/null && pwd)"
 cd "${CUR_DIR}"
 
 CONAN_CENTER_COMMIT_ID="bad5c95b810e859c1c31553b92584246fe436d69"
-CCI_HOME="${CONAN_HOME:-~/.conan2}/conan-center-index"
+CCI_HOME="${CONAN_HOME:-$HOME/.conan2}/conan-center-index"
 
 if ! command -v conan &> /dev/null; then
   echo "❌ Error: 'conan' command not found."