Move Dockerfile logic into extconf.rb

This change is in preparation for building a proper gem that can be
installed from RubyGems.

Dockerfile

@@ -32,29 +32,11 @@ RUN wget -q -O $CLANG_FILE $CLANG_URL && \
     tar xf $CLANG_FILE -C $CLANG_DIR --strip-components 1 && \
     rm $CLANG_FILE
 
-# https://github.com/google/sanitizers/wiki/AddressSanitizerFlags
-ENV CC="$CLANG_DIR/bin/clang"
-ENV CXX="$CLANG_DIR/bin/clang++"
-ENV LDSHARED="$CLANG_DIR/bin/clang -shared"
-ENV LDSHAREDXX="$CLANG_DIR/bin/clang++ -shared"
-ENV ASAN_SYMBOLIZER_PATH="$CLANG_DIR/bin/llvm-symbolizer"
 ENV PATH="$PATH:$CLANG_DIR/bin"
-
-ENV FUZZER_NO_MAIN_LIB="$CLANG_DIR/lib/clang/17/lib/$CLANG_ARCH-unknown-linux-gnu/libclang_rt.fuzzer_no_main.a"
-ENV ASAN_LIB="$CLANG_DIR/lib/clang/17/lib/$CLANG_ARCH-unknown-linux-gnu/libclang_rt.asan.a"
-ENV ASAN_STRIPPED_LIB="/tmp/libclang_rt.asan.a"
-ENV ASAN_MERGED_LIB="/tmp/asan_with_fuzzer.so"
-
-# https://github.com/google/atheris/blob/master/native_extension_fuzzing.md#why-this-is-necessary
-RUN cp "$ASAN_LIB" "$ASAN_STRIPPED_LIB"
-RUN ar d "$ASAN_STRIPPED_LIB" asan_preinit.cc.o asan_preinit.cpp.o
-RUN "$CXX" \
-    -Wl,--whole-archive \
-    "$FUZZER_NO_MAIN_LIB" \
-    "$ASAN_STRIPPED_LIB" \
-    -Wl,--no-whole-archive \
-    -lpthread -ldl -shared \
-    -o "$ASAN_MERGED_LIB"
+ENV CC="clang"
+ENV CXX="clang++"
+ENV LDSHARED="clang -shared"
+ENV LDSHAREDXX="clang++ -shared"
 
 # The MAKE variable allows overwriting the make command at runtime. This forces the
 # Ruby C extension to respect ENV variables when compiling, like CC, CFLAGS, etc.
@@ -71,7 +53,8 @@ WORKDIR ruzzy/
 RUN bundler3.1 install
 
 COPY . .
-RUN rake compile
+RUN gem build
+RUN gem install --verbose ruzzy-*.gem
 
 ENTRYPOINT ["./entrypoint.sh"]
 CMD ["-help=1"]

Rakefile

@@ -4,15 +4,18 @@ require 'rake/testtask'
 require 'rake/extensiontask'
 
 Rake::TestTask.new do |t|
+  require 'ruzzy'
+
   # This is required for tests that use cruzzy functionality
-  ENV['LD_PRELOAD'] = ENV['ASAN_MERGED_LIB']
+  ENV['LD_PRELOAD'] = Ruzzy.ext_path + '/' + 'asan_with_fuzzer.so'
+
   t.verbose = true
 end
 
 Rake::ExtensionTask.new 'cruzzy' do |ext|
   ext.lib_dir = 'lib/cruzzy'
 end
 
-Rake::ExtensionTask.new 'cruzzy/dummy' do |ext|
-  ext.lib_dir = 'lib/cruzzy/dummy'
+Rake::ExtensionTask.new 'dummy' do |ext|
+  ext.lib_dir = 'lib/dummy'
 end

bin/dummy.rb

@@ -2,6 +2,6 @@
 
 require 'ruzzy'
 
-dummy_test_one_input = ->(data) { Ruzzy.c_dummy_test_one_input(data) }
+dummy_test_one_input = ->(data) { Ruzzy.dummy_test_one_input(data) }
 
 Ruzzy.fuzz(dummy_test_one_input)

entrypoint.sh

@@ -1,5 +1,4 @@
 #!/bin/bash
 
-export LD_PRELOAD=${ASAN_MERGED_LIB}
-
-ruby -Ilib bin/dummy.rb "$@"
+LD_PRELOAD=$(ruby -e 'require "ruzzy"; print Ruzzy.ext_path')/asan_with_fuzzer.so \
+    ruby bin/dummy.rb "$@"

ext/cruzzy/extconf.rb

@@ -2,26 +2,80 @@
 
 require 'mkmf'
 require 'open3'
+require 'tempfile'
 
-lib_name = 'libclang_rt.fuzzer_no_main.a'
-env_name = 'FUZZER_NO_MAIN_LIB'
-clang_name = 'clang'
+CLANG = 'clang'
+FUZZER_NO_MAIN_LIB = 'FUZZER_NO_MAIN_LIB'
 
-lib_path, status = Open3.capture2(clang_name, "--print-file-name", lib_name)
+find_executable(CLANG)
 
-if !status.success?
-  puts("Could not find #{lib_name} using #{clang_name}.")
-  lib_path = ENV.fetch(env_name, nil)
-  if lib_path.nil?
-    puts("Could not find #{lib_name} in #{env_name}.")
-    puts("Please include #{clang_name} in your path or specify #{env_name} ENV variable.")
+def get_clang_file_name(file_name)
+  stdout, status = Open3.capture2(CLANG, '--print-file-name', file_name)
+  status.success? && File.exist?(stdout.strip) ? stdout.strip : false
+end
+
+def merge_asan_libfuzzer_lib(asan_lib, fuzzer_no_main_lib)
+  # https://github.com/google/atheris/blob/master/native_extension_fuzzing.md#why-this-is-necessary
+  Tempfile.create do |file|
+    file.write(File.open(asan_lib).read)
+
+    stdout, status = Open3.capture2(
+      'ar',
+      'd',
+      file.path,
+      'asan_preinit.cc.o',
+      'asan_preinit.cpp.o'
+    )
+
+    stdout, status = Open3.capture2(
+      ENV['CXX'],
+      '-Wl,--whole-archive',
+      fuzzer_no_main_lib,
+      file.path,
+      '-Wl,--no-whole-archive',
+      '-lpthread',
+      '-ldl',
+      '-shared',
+      '-o',
+      'asan_with_fuzzer.so'
+    )
+  end
+end
+
+fuzzer_no_main_libs = [
+  'libclang_rt.fuzzer_no_main.a',
+  'libclang_rt.fuzzer_no_main-aarch64.a',
+  'libclang_rt.fuzzer_no_main-x86_64.a'
+]
+fuzzer_no_main_lib = fuzzer_no_main_libs.map { |lib| get_clang_file_name(lib) }.find(&:itself)
+
+unless fuzzer_no_main_lib
+  puts("Could not find fuzzer_no_main using #{CLANG}.")
+  fuzzer_no_main_lib = ENV.fetch(FUZZER_NO_MAIN_LIB, nil)
+  if fuzzer_no_main_lib.nil?
+    puts("Could not find fuzzer_no_main in #{FUZZER_NO_MAIN_LIB}.")
+    puts("Please include #{CLANG} in your path or specify #{FUZZER_NO_MAIN_LIB} ENV variable.")
     exit(1)
   end
 end
 
 # The LOCAL_LIBS variable allows linking arbitrary libraries into Ruby C
 # extensions. It is supported by the Ruby mkmf library and C extension Makefile.
 # For more information, see https://github.com/ruby/ruby/blob/master/lib/mkmf.rb.
-$LOCAL_LIBS = lib_path
+$LOCAL_LIBS = fuzzer_no_main_lib
+
+asan_libs = [
+  'libclang_rt.asan.a',
+  'libclang_rt.asan-aarch64.a',
+  'libclang_rt.asan-x86_64.a'
+]
+asan_lib = asan_libs.map { |lib| get_clang_file_name(lib) }.find(&:itself)
+
+unless asan_lib
+  puts("Could not find asan using #{CLANG}.")
+  exit(1)
+end
+
+merge_asan_libfuzzer_lib(asan_lib, fuzzer_no_main_lib)
 
 create_makefile('cruzzy/cruzzy')

ext/cruzzy/dummy/extconf.rb → ext/dummy/extconf.rb

@@ -2,7 +2,8 @@
 
 require 'mkmf'
 
+# https://github.com/google/sanitizers/wiki/AddressSanitizerFlags
 $CFLAGS = '-fsanitize=address,fuzzer-no-link -fno-omit-frame-pointer -fno-common -fPIC -g'
 $CXXFLAGS = '-fsanitize=address,fuzzer-no-link -fno-omit-frame-pointer -fno-common -fPIC -g'
 
-create_makefile('cruzzy/cruzzy/dummy')
+create_makefile('dummy/dummy')

lib/ruzzy.rb

@@ -1,15 +1,31 @@
 # frozen_string_literal: true
 
+require 'pathname'
+
 # A Ruby C extension fuzzer
 module Ruzzy
   require 'cruzzy/cruzzy'
-  require 'cruzzy/dummy/dummy'
 
   DEFAULT_ARGS = [$PROGRAM_NAME] + ARGV
 
   def fuzz(test_one_input, args = DEFAULT_ARGS)
     c_fuzz(test_one_input, args)
   end
 
+  def ext_path
+    (Pathname.new(__FILE__).parent.parent + 'ext' + 'cruzzy').to_s
+  end
+
+  def dummy_test_one_input(data)
+    # This 'require' depends on LD_PRELOAD, so it's placed inside the function
+    # scope. This allows us to run ext_path for LD_PRELOAD and not have a
+    # circular dependency.
+    require 'dummy/dummy'
+
+    c_dummy_test_one_input(data)
+  end
+
   module_function :fuzz
+  module_function :ext_path
+  module_function :dummy_test_one_input
 end

ruzzy.gemspec

@@ -1,17 +1,15 @@
 # frozen_string_literal: true
 
-require 'rake'
-
 Gem::Specification.new do |s|
   s.name                  = 'ruzzy'
   s.version               = '0.5.0'
   s.summary               = 'A Ruby C extension fuzzer'
   s.authors               = ['Trail of Bits']
   s.email                 = '[email protected]'
-  s.files                 = FileList['lib/**/*.rb', 'ext/**/*.{rb,c,h}']
+  s.files                 = Dir['lib/**/*.rb'] + Dir['ext/**/*.{rb,c,h}']
   s.homepage              = 'https://rubygems.org/gems/ruzzy'
   s.license               = 'AGPL-3.0'
-  s.extensions            = %w[ext/cruzzy/extconf.rb]
+  s.extensions            = %w[ext/cruzzy/extconf.rb ext/dummy/extconf.rb]
   s.required_ruby_version = '>= 3.1.0'
 
   s.add_development_dependency 'rake', '~> 13.0'

test/test_ruzzy.rb

@@ -10,29 +10,29 @@ def test_c_libfuzzer_is_loaded
     assert_true(result)
   end
 
-  def test_c_dummy_test_one_input_proc
-    dummy_test_one_input = proc { |data| Ruzzy.c_dummy_test_one_input(data) }
+  def test_dummy_test_one_input_proc
+    dummy_test_one_input = proc { |data| Ruzzy.dummy_test_one_input(data) }
 
     result = dummy_test_one_input.call('test')
     expected = 0
 
     assert_equal(result, expected)
   end
 
-  def test_c_dummy_test_one_input_lambda
-    dummy_test_one_input = ->(data) { Ruzzy.c_dummy_test_one_input(data) }
+  def test_dummy_test_one_input_lambda
+    dummy_test_one_input = ->(data) { Ruzzy.dummy_test_one_input(data) }
 
     result = dummy_test_one_input.call('test')
     expected = 0
 
     assert_equal(result, expected)
   end
 
-  def test_c_dummy_test_one_input_invalid_return
+  def test_dummy_test_one_input_invalid_return
     omit("This test calls LLVMFuzzerRunDriver, which we don't have a good harness for yet")
 
     dummy_test_one_input = lambda do |data|
-      Ruzzy.c_dummy_test_one_input(data)
+      Ruzzy.dummy_test_one_input(data)
       'not an integer or nil'
     end
 
@@ -46,4 +46,8 @@ def test_fuzz_without_proc
       Ruzzy.fuzz('not a proc')
     end
   end
+
+  def test_ext_path
+    assert(Ruzzy.ext_path)
+  end
 end