[sanitizer] Add CMake flag to build with internal symbolizer

This intermediate result in moving internal symbolizer build
from sh script to CMake rules.

The flag is supposed to be used with:
-DLLVM_ENABLE_PROJECTS="clang;lld;compiler-rt" -DLLVM_ENABLE_RUNTIMES="libunwind;libcxx;libcxxabi" -Sllvm-project/llvm

After converting sh script into cmake, we may add support for other build modes.

For https://github.com/llvm/llvm-project/issues/30098

Reviewed By: kstoimenov, MaskRay

Differential Revision: https://reviews.llvm.org/D157947

compiler-rt/CMakeLists.txt

@@ -727,6 +727,9 @@ endif()
 pythonize_bool(COMPILER_RT_HAS_LLD)
 pythonize_bool(COMPILER_RT_TEST_USE_LLD)
 
+option(COMPILER_RT_ENABLE_INTERNAL_SYMBOLIZER "Build Compiler RT linked with in LLVM symbolizer" OFF)
+mark_as_advanced(COMPILER_RT_ENABLE_INTERNAL_SYMBOLIZER)
+
 add_subdirectory(lib)
 
 if(COMPILER_RT_INCLUDE_TESTS)

compiler-rt/lib/sanitizer_common/CMakeLists.txt

@@ -362,3 +362,7 @@ foreach(arch ${SANITIZER_COMMON_SUPPORTED_ARCH})
   add_library(RTSanitizerCommonSymbolizerInternal.${arch}
               OBJECT IMPORTED GLOBAL)
 endforeach()
+
+if (COMPILER_RT_ENABLE_INTERNAL_SYMBOLIZER)
+  add_subdirectory(symbolizer)
+endif()

compiler-rt/lib/sanitizer_common/symbolizer/CMakeLists.txt

@@ -0,0 +1,27 @@
+foreach(arch ${SANITIZER_COMMON_SUPPORTED_ARCH})
+  get_target_flags_for_arch(${arch} TARGET_CFLAGS)
+
+  set(RTSanitizerCommonSymbolizerInternalDir
+      "${CMAKE_CURRENT_BINARY_DIR}/RTSanitizerCommonSymbolizerInternal.${arch}")
+  add_custom_command(OUTPUT ${RTSanitizerCommonSymbolizerInternalDir}
+                     COMMAND ${CMAKE_COMMAND} -E make_directory ${RTSanitizerCommonSymbolizerInternalDir})
+
+  add_custom_command(OUTPUT RTSanitizerCommonSymbolizerInternal.${arch}.o
+                     DEPENDS ${RTSanitizerCommonSymbolizerInternalDir} 
+                             clang lld llvm-tblgen opt llvm-ar llvm-link llvm-ranlib llvm-symbolizer
+                             sanitizer_wrappers.cpp
+                             sanitizer_symbolize.cpp
+                             scripts/build_symbolizer.sh
+                     WORKING_DIRECTORY ${RTSanitizerCommonSymbolizerInternalDir}
+                     COMMAND FLAGS=${TARGET_CFLAGS}
+                             CLANG=${LLVM_RUNTIME_OUTPUT_INTDIR}/clang
+                             ${CMAKE_CURRENT_SOURCE_DIR}/scripts/build_symbolizer.sh
+                             ${CMAKE_CURRENT_BINARY_DIR}/RTSanitizerCommonSymbolizerInternal.${arch}.o
+                     USES_TERMINAL)
+
+  add_custom_target(RTSanitizerCommonSymbolizerInternalObj.${arch}
+                    DEPENDS RTSanitizerCommonSymbolizerInternal.${arch}.o)
+
+  set_property(TARGET RTSanitizerCommonSymbolizerInternal.${arch}
+               PROPERTY IMPORTED_OBJECTS ${CMAKE_CURRENT_BINARY_DIR}/RTSanitizerCommonSymbolizerInternal.${arch}.o)
+endforeach()

compiler-rt/lib/sanitizer_common/symbolizer/scripts/build_symbolizer.sh

@@ -3,26 +3,16 @@
 # Run as: CLANG=bin/clang build_symbolizer.sh out.o
 # zlib can be downloaded from http://www.zlib.net.
 #
-# Script compiles self-contained object file with symbolization code and injects
-# it into the given set of runtime libraries. Script updates only libraries
-# which has unresolved __sanitizer_symbolize_* symbols and matches architecture.
-# Object file is be compiled from LLVM sources with dependencies like libc++ and
-# zlib. Then it internalizes symbols in the file, so that it can be linked
-# into arbitrary programs, avoiding conflicts with the program own symbols and
-# avoiding dependencies on any program symbols. The only acceptable dependencies
-# are libc and __sanitizer::internal_* from sanitizer runtime.
+# Script compiles self-contained object file with symbolization code.
 #
 # Symbols exported by the object file will be used by Sanitizer runtime
 # libraries to symbolize code/data in-process.
 #
-# The script will modify the output directory which is given as the first
-# argument to the script.
-#
 # FIXME: We should really be using a simpler approach to building this object
 # file, and it should be available as a regular cmake rule. Conceptually, we
 # want to be doing "ld -r" followed by "objcopy -G" to create a relocatable
 # object file with only our entry points exposed. However, this does not work at
-# present, see PR30750.
+# present, see https://github.com/llvm/llvm-project/issues/30098.
 
 set -x
 set -e
@@ -30,7 +20,7 @@ set -u
 
 SCRIPT_DIR=$(cd "$(dirname "$0")" && pwd)
 SRC_DIR=$(readlink -f $SCRIPT_DIR/..)
-TARGE_DIR=$(readlink -f $1)
+OUTPUT=$(readlink -f $1)
 COMPILER_RT_SRC=$(readlink -f ${SCRIPT_DIR}/../../../..)
 LLVM_SRC=${LLVM_SRC:-${COMPILER_RT_SRC}/../llvm}
 LLVM_SRC=$(readlink -f $LLVM_SRC)
@@ -186,20 +176,6 @@ nm -f posix -g symbolizer.o | cut -f 1,2 -d \  | LC_COLLATE=C sort -u > undefine
 (diff -u $SCRIPT_DIR/global_symbols.txt undefined.new | grep -E "^\+[^+]") && \
   (echo "Failed: unexpected symbols"; exit 1)
 
-arch() {
-  objdump -f $1 | grep -m1 -Po "(?<=file format ).*$"
-}
-
-SYMBOLIZER_FORMAT=$(arch symbolizer.o)
-echo "Injecting $SYMBOLIZER_FORMAT symbolizer..."
-for A in $TARGE_DIR/libclang_rt.*san*.a; do
-  A_FORMAT=$(arch $A)
-  if [[ "$A_FORMAT" != "$SYMBOLIZER_FORMAT" ]] ; then
-    continue
-  fi
-  (nm -u $A 2>/dev/null | grep -E "__sanitizer_symbolize_code" >/dev/null) || continue
-  echo "$A"
-  $AR rcs $A symbolizer.o
-done
+cp -f symbolizer.o $OUTPUT
 
 echo "Success!"