[cmake] Add support for statically linking libxml2

Dynamically depending on libxml2 results in various annoyances across
different linux distros for release artifacts. Specifically on fedora
and nixos the library has a different name than on debian, and on
arch-linux they tried to remove the old name entirely.

With this, enabled by default for releases, we don't sacrifice any
behavior changes, but no longer have these issues. For lld the binary
size impact is <1mb

macOS ignores this setting since libxml2 is part of the OS and stable
enough.

This mirrors what we do for zstd

Fixes #113696
Fixes #138225
Fixes https://discourse.llvm.org/t/official-builds-without-libxml2-and-libtinfo/58169

Author: keith

clang/cmake/caches/Release.cmake

@@ -171,6 +171,7 @@ set_final_stage_var(CPACK_GENERATOR "TXZ" STRING)
 set_final_stage_var(CPACK_ARCHIVE_THREADS "0" STRING)
 
 set_final_stage_var(LLVM_USE_STATIC_ZSTD "ON" BOOL)
+set_final_stage_var(LLVM_USE_STATIC_LIBXML2 "ON" BOOL)
 if (LLVM_RELEASE_ENABLE_LTO)
   set_final_stage_var(LLVM_ENABLE_FATLTO "ON" BOOL)
   set_final_stage_var(CPACK_PRE_BUILD_SCRIPTS "${CMAKE_CURRENT_LIST_DIR}/release_cpack_pre_build_strip_lto.cmake" STRING)

lldb/source/Host/CMakeLists.txt

@@ -152,7 +152,7 @@ set(EXTRA_LIBS)
 if (CMAKE_SYSTEM_NAME MATCHES "NetBSD")
   list(APPEND EXTRA_LIBS kvm)
 endif()
-if (LLDB_ENABLE_LIBXML2)
+if (LLDB_ENABLE_LIBXML2 AND NOT LLVM_USE_STATIC_LIBXML2)
   list(APPEND EXTRA_LIBS LibXml2::LibXml2)
 endif()
 if (HAVE_LIBDL)
@@ -190,3 +190,8 @@ add_lldb_library(lldbHost NO_PLUGIN_DEPENDENCIES
     ${LLDB_LIBEDIT_LIBS}
   )
 
+if (LLDB_ENABLE_LIBXML2 AND LLVM_USE_STATIC_LIBXML2)
+  target_link_libraries(lldbHost PRIVATE ${LIBXML2_LIBRARIES})
+  target_include_directories(lldbHost PUBLIC ${LIBXML2_INCLUDE_DIRS})
+  add_dependencies(lldbHost libxml2)
+endif()

llvm/CMakeLists.txt

@@ -605,6 +605,8 @@ set(LLVM_TARGET_ARCH "host"
 
 set(LLVM_ENABLE_LIBXML2 "ON" CACHE STRING "Use libxml2 if available. Can be ON, OFF, or FORCE_ON")
 
+set(LLVM_USE_STATIC_LIBXML2 "OFF" CACHE BOOL "Use static version of libxml2. Can be ON, or OFF")
+
 option(LLVM_ENABLE_LIBEDIT "Use libedit if available." ON)
 
 option(LLVM_ENABLE_LIBPFM "Use libpfm for performance counters if available." ON)

llvm/cmake/config-ix.cmake

@@ -210,22 +210,31 @@ if(LLVM_ENABLE_ZSTD)
 endif()
 
 if(LLVM_ENABLE_LIBXML2)
-  if(LLVM_ENABLE_LIBXML2 STREQUAL FORCE_ON)
-    find_package(LibXml2 REQUIRED)
-  elseif(NOT LLVM_USE_SANITIZER MATCHES "Memory.*")
-    find_package(LibXml2)
+  if(APPLE)
+    set(LLVM_USE_STATIC_LIBXML2 OFF)
   endif()
-  if(LibXml2_FOUND)
-    # Check if libxml2 we found is usable; for example, we may have found a 32-bit
-    # library on a 64-bit system which would result in a link-time failure.
-    cmake_push_check_state()
-    list(APPEND CMAKE_REQUIRED_INCLUDES ${LIBXML2_INCLUDE_DIRS})
-    list(APPEND CMAKE_REQUIRED_LIBRARIES ${LIBXML2_LIBRARIES})
-    list(APPEND CMAKE_REQUIRED_DEFINITIONS ${LIBXML2_DEFINITIONS})
-    check_symbol_exists(xmlReadMemory libxml/xmlreader.h HAVE_LIBXML2)
-    cmake_pop_check_state()
-    if(LLVM_ENABLE_LIBXML2 STREQUAL FORCE_ON AND NOT HAVE_LIBXML2)
-      message(FATAL_ERROR "Failed to configure libxml2")
+
+  if(LLVM_USE_STATIC_LIBXML2)
+    include(LibXml2)
+    set(HAVE_LIBXML2 1)
+  else()
+    if(LLVM_ENABLE_LIBXML2 STREQUAL FORCE_ON)
+      find_package(LibXml2 REQUIRED)
+    elseif(NOT LLVM_USE_SANITIZER MATCHES "Memory.*")
+      find_package(LibXml2)
+    endif()
+    if(LibXml2_FOUND)
+      # Check if libxml2 we found is usable; for example, we may have found a 32-bit
+      # library on a 64-bit system which would result in a link-time failure.
+      cmake_push_check_state()
+      list(APPEND CMAKE_REQUIRED_INCLUDES ${LIBXML2_INCLUDE_DIRS})
+      list(APPEND CMAKE_REQUIRED_LIBRARIES ${LIBXML2_LIBRARIES})
+      list(APPEND CMAKE_REQUIRED_DEFINITIONS ${LIBXML2_DEFINITIONS})
+      check_symbol_exists(xmlReadMemory libxml/xmlreader.h HAVE_LIBXML2)
+      cmake_pop_check_state()
+      if(LLVM_ENABLE_LIBXML2 STREQUAL FORCE_ON AND NOT HAVE_LIBXML2)
+        message(FATAL_ERROR "Failed to configure libxml2")
+      endif()
     endif()
   endif()
   set(LLVM_ENABLE_LIBXML2 "${HAVE_LIBXML2}")

llvm/cmake/modules/LibXml2.cmake

@@ -0,0 +1,31 @@
+include(ExternalProject)
+
+if (NOT LIBXML2_PREFIX)
+  set (LIBXML2_PREFIX libxml2)
+endif()
+
+set(LIBXML2_PATH ${CMAKE_CURRENT_BINARY_DIR}/${LIBXML2_PREFIX}/src/${LIBXML2_PREFIX})
+set(LIBXML2_LIB_PATH ${LIBXML2_PATH}-build/libxml2.a)
+
+ExternalProject_Add(${LIBXML2_PREFIX}
+  PREFIX ${LIBXML2_PREFIX}
+  GIT_REPOSITORY https://github.com/GNOME/libxml2.git
+  GIT_TAG v2.15.1
+  CMAKE_ARGS -DCMAKE_BUILD_TYPE=${CMAKE_BUILD_TYPE}
+             -DBUILD_SHARED_LIBS=OFF
+             -DLIBXML2_WITH_PYTHON=OFF
+             -DLIBXML2_WITH_PROGRAMS=OFF
+             -DLIBXML2_WITH_TESTS=OFF
+             -DLIBXML2_WITH_LZMA=OFF
+             -DLIBXML2_WITH_ZLIB=OFF
+  CMAKE_CACHE_ARGS -DCMAKE_C_COMPILER:FILEPATH=${CMAKE_C_COMPILER}
+                   -DCMAKE_CXX_COMPILER:FILEPATH=${CMAKE_CXX_COMPILER}
+                   -DCMAKE_POSITION_INDEPENDENT_CODE:BOOL=ON
+  BUILD_BYPRODUCTS ${LIBXML2_LIB_PATH}
+  UPDATE_COMMAND ""
+  INSTALL_COMMAND ""
+  )
+
+set(LIBXML2_INCLUDE_DIRS ${LIBXML2_PATH}/include ${LIBXML2_PATH}-build)
+set(LIBXML2_LIBRARIES ${LIBXML2_LIB_PATH})
+set(LIBXML2_DEFINITIONS "")

llvm/lib/WindowsManifest/CMakeLists.txt

@@ -1,6 +1,6 @@
 include(GetLibraryName)
 
-if(LLVM_ENABLE_LIBXML2)
+if(LLVM_ENABLE_LIBXML2 AND NOT LLVM_USE_STATIC_LIBXML2)
   set(imported_libs LibXml2::LibXml2)
 endif()
 
@@ -18,17 +18,28 @@ add_llvm_component_library(LLVMWindowsManifest
   Support
   )
 
+if(LLVM_ENABLE_LIBXML2 AND LLVM_USE_STATIC_LIBXML2)
+  target_link_libraries(LLVMWindowsManifest PRIVATE ${LIBXML2_LIBRARIES})
+  target_include_directories(LLVMWindowsManifest PRIVATE ${LIBXML2_INCLUDE_DIRS})
+  add_dependencies(LLVMWindowsManifest libxml2)
+endif()
+
 # This block is only needed for llvm-config. When we deprecate llvm-config and
 # move to using CMake export, this block can be removed.
 if(LLVM_ENABLE_LIBXML2)
-  # CMAKE_BUILD_TYPE is only meaningful to single-configuration generators.
-  if(CMAKE_BUILD_TYPE)
-    string(TOUPPER ${CMAKE_BUILD_TYPE} build_type)
-    get_property(libxml2_library TARGET LibXml2::LibXml2 PROPERTY LOCATION_${build_type})
-  endif()
-  if(NOT libxml2_library)
-    get_property(libxml2_library TARGET LibXml2::LibXml2 PROPERTY LOCATION)
+  if(LLVM_USE_STATIC_LIBXML2)
+    # When using static libxml2 built via ExternalProject, use the library path directly
+    get_library_name(${LIBXML2_LIBRARIES} libxml2_library)
+  else()
+    # CMAKE_BUILD_TYPE is only meaningful to single-configuration generators.
+    if(CMAKE_BUILD_TYPE)
+      string(TOUPPER ${CMAKE_BUILD_TYPE} build_type)
+      get_property(libxml2_library TARGET LibXml2::LibXml2 PROPERTY LOCATION_${build_type})
+    endif()
+    if(NOT libxml2_library)
+      get_property(libxml2_library TARGET LibXml2::LibXml2 PROPERTY LOCATION)
+    endif()
+    get_library_name(${libxml2_library} libxml2_library)
   endif()
-  get_library_name(${libxml2_library} libxml2_library)
   set_property(TARGET LLVMWindowsManifest PROPERTY LLVM_SYSTEM_LIBS ${libxml2_library})
 endif()