[libc] [startup] add cmake function to merge separated crt1 objects

[SchrodingerZhu]

As part of startup refactoring, this patch adds a function to merge multiple objects into a single relocatable object:
cc -r obj1.o obj2.o -o obj.o

A relocatable object is an object file that is not fully linked into an executable or a shared library. It is an intermediate file format that can be passed into the linker.

A crt object can have arch-specific code and arch-agnostic code. To reduce code cohesion, the implementation is splitted into multiple units. As a result, we need to merge them into a single relocatable object.

[llvmbot]

@llvm/pr-subscribers-libc

Author: Schrodinger ZHU Yifan (SchrodingerZhu)

Changes

As part of startup refactoring, this patch adds a function to merge multiple objects into a single relocatable object:
cc -r obj1.o obj2.o -o obj.o

A relocatable object is an object file that is not fully linked into an executable or a shared library. It is an intermediate file format that can be passed into the linker.

A crt object can have arch-specific code and arch-agnostic code. To reduce code cohesion, the implementation is splitted into multiple units. As a result, we need to merge them into a single relocatable object.

---

Full diff: https://github.com/llvm/llvm-project/pull/75413.diff

1 Files Affected:

• (modified) libc/startup/linux/CMakeLists.txt (+42-4)

diff --git a/libc/startup/linux/CMakeLists.txt b/libc/startup/linux/CMakeLists.txt
index 007aa30c17d6ab..5ca5010f6b1fe4 100644
--- a/libc/startup/linux/CMakeLists.txt
+++ b/libc/startup/linux/CMakeLists.txt
@@ -1,3 +1,42 @@
+# This function merges multiple objects into a single relocatable object
+#                     cc -r obj1.o obj2.o -o obj.o
+# A relocatable object is an object file that is not fully linked into an
+# executable or a shared library. It is an intermediate file format that can
+# be passed into the linker.
+# A crt object have arch-specific code and arch-agnostic code. To reduce code
+# cohesion, the implementation is splitted into multiple units. As a result,
+# we need to merge them into a single relocatable object.
+function(merge_relocatable_object name)
+  set(obj_list "")
+  set(fq_link_libraries "")
+  get_fq_deps_list(fq_dep_list ${ARGN})
+  foreach(target IN LISTS fq_dep_list)
+    list(APPEND obj_list "$<TARGET_OBJECTS:${target}>")
+    get_target_property(libs ${target} DEPS)
+    list(APPEND fq_link_libraries "${libs}")
+  endforeach()
+  get_fq_target_name(${name} fq_name)
+  add_custom_command(
+    OUTPUT ${CMAKE_CURRENT_BINARY_DIR}/${name}.o
+    COMMAND ${CMAKE_CXX_COMPILER} -r ${obj_list} -o ${CMAKE_CURRENT_BINARY_DIR}/${name}.o
+    DEPENDS ${obj_list}
+    COMMAND_EXPAND_LISTS
+  )
+  add_custom_target(${fq_name}.relocatable DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/${name}.o)
+  add_library(${fq_name} OBJECT IMPORTED GLOBAL)
+  add_dependencies(${fq_name} ${fq_name}.relocatable)
+  target_link_libraries(${fq_name} INTERFACE ${fq_link_libraries})
+  set_target_properties(
+    ${fq_name} 
+    PROPERTIES
+      LINKER_LANGUAGE CXX
+      OBJECT_FILES ${CMAKE_CURRENT_BINARY_DIR}/${name}.o
+      IMPORTED_OBJECTS ${CMAKE_CURRENT_BINARY_DIR}/${name}.o
+      TARGET_TYPE ${OBJECT_LIBRARY_TARGET_TYPE}
+      DEPS "${fq_link_libraries}"
+  ) 
+endfunction()
+
 function(add_startup_object name)
   cmake_parse_arguments(
     "ADD_STARTUP_OBJECT"
@@ -34,11 +73,10 @@ endif()
 
 add_subdirectory(${LIBC_TARGET_ARCHITECTURE})
 
-add_startup_object(
+# TODO: factor out crt1 into multiple objects
+merge_relocatable_object(
   crt1
-  ALIAS
-  DEPENDS
-    .${LIBC_TARGET_ARCHITECTURE}.crt1
+  .${LIBC_TARGET_ARCHITECTURE}.crt1
 )
 
 add_startup_object(

[SchrodingerZhu]

@jhuber6 unfortunately, llvm_check_linker_flag does not really work on my computer (I thought it had worked but it was due to the cache). However, check_cxx_compiler_flag("-r" LIBC_LINKER_SUPPORTS_RELOCATABLE) works fine.

[SchrodingerZhu]

Call for a merge if no further objection.

[nickdesaulniers]

ah, @SchrodingerZhu it looks like this broke the post submit bots:

• asan full build: https://lab.llvm.org/buildbot/#/builders/162
• aarch64: https://lab.llvm.org/buildbot/#/builders/223

[nickdesaulniers]

FAILED: projects/libc/startup/linux/crt1.o 
: && /usr/bin/clang++ -fPIC -fvisibility-inlines-hidden -Werror=date-time -Werror=unguarded-availability-new -Wall -Wextra -Wno-unused-parameter -Wwrite-strings -Wcast-qual -Wmissing-field-initializers -pedantic -Wno-long-long -Wc++98-compat-extra-semi -Wimplicit-fallthrough -Wcovered-switch-default -Wno-noexcept-type -Wnon-virtual-dtor -Wdelete-non-virtual-dtor -Wsuggest-override -Wstring-conversion -Wmisleading-indentation -Wctad-maybe-unsupported -fno-omit-frame-pointer -O1 -fsanitize=address -fdiagnostics-color -g -r projects/libc/startup/linux/x86_64/CMakeFiles/libc.startup.linux.x86_64.crt1.dir/start.cpp.o -o projects/libc/startup/linux/crt1.o   && :
/usr/bin/ld: projects/libc/startup/linux/x86_64/CMakeFiles/libc.startup.linux.x86_64.crt1.dir/start.cpp.o: in function `_start':
/home/llvm-libc-buildbot/buildbot-worker/libc-x86_64-debian-fullbuild/libc-x86_64-debian-fullbuild-dbg-asan/llvm-project/libc/startup/linux/x86_64/start.cpp:144: multiple definition of `_start'; /usr/bin/../lib/gcc/x86_64-linux-gnu/8/../../../x86_64-linux-gnu/crt1.o:(.text+0x0): first defined here

So it looks like _start is multiply defined:

• crt1.o (from the system)
• start.cpp.o (from llvmlibc)

Can you see if this can be fixed forward quickly? Otherwise I can revert.

I suspect -DLLVM_LIBC_FULL_BUILD=ON are the broken builds.

[SchrodingerZhu]

I actually had full build on locally. Inspecting the log, I think it is due to the unaligned behavior of the compilers. I suppose adding nostartfiles or freestanding should address the problem.

[SchrodingerZhu]

@nickdesaulniers maybe revert it for now? I cannot work on this right now. I can get to this 2-3 hours later.

[SchrodingerZhu]

Never mind. My meeting was canceled just now. I can submit a quick fix attempt with -nostdlib.