Add self contained builder

Author: boomanaiden154

llvm_ir_dataset_utils/builders/builder.py

@@ -14,7 +14,8 @@
 from llvm_ir_dataset_utils.builders import (autoconf_builder, cargo_builder,
                                             cmake_builder, julia_builder,
                                             manual_builder, spack_builder,
-                                            swift_builder, portage_builder)
+                                            swift_builder, portage_builder,
+                                            self_contained_builder)
 from llvm_ir_dataset_utils.sources import source
 from llvm_ir_dataset_utils.util import file, licenses
 
@@ -235,6 +236,10 @@ def parse_and_build_from_description(
         threads,
         corpus_description["package_name"],
     )
+  elif corpus_description["build_system"] == "self_contained":
+    build_log = self_contained_builder.perform_build(
+        corpus_description["source_file_list"], build_dir, corpus_dir)
+    self_contained_builder.extract_ir(build_dir, corpus_dir, threads)
   else:
     raise ValueError(
         f"Build system {corpus_description['build_system']} is not supported")

llvm_ir_dataset_utils/builders/self_contained_builder.py

@@ -0,0 +1,43 @@
+"""Module for building and extracting builder from a set of self-contained
+C/c++ files."""
+
+import subprocess
+import os
+
+from mlgo.corpus import extract_ir_lib
+from mlgo.corpus import make_corpus_lib
+
+
+def compile_file(source_file, object_file):
+  command_vector = [
+      'clang', '-Xclang', '-fembed-bitcode=all', source_file, '-o', object_file
+  ]
+  compile_process = subprocess.run(
+      command_vector, stderr=subprocess.STDOUT, stdout=subprocess.PIPE)
+  assert (compile_process.returncode == 0)
+
+
+def perform_build(source_file_list, build_dir, corpus_dir):
+  for source_file in source_file_list:
+    object_file = os.path.join(build_dir, os.path.basename(source_file) + '.o')
+    compile_file(source_file, object_file)
+
+  return {
+      'targets': [{
+          'success': True,
+          'build_log': None,
+          'name': 'self_contained'
+      }]
+  }
+
+
+# TODO(boomanaiden154): This is duplicated with extract_ir in the manual builder.
+# We might want to look into refactoring to consolidate the two functions at some
+# point.
+def extract_ir(build_dir, corpus_dir, threads):
+  objects = extract_ir_lib.load_from_directory(build_dir, corpus_dir)
+  relative_output_paths = extract_ir_lib.run_extraction(objects, threads,
+                                                        "llvm-objcopy", None,
+                                                        None, ".llvmcmd",
+                                                        ".llvmbc")
+  extract_ir_lib.write_corpus_manifest(None, relative_output_paths, corpus_dir)