build: bring abseil submodule back

because of https://bugzilla.redhat.com/show_bug.cgi?id=2278689,
the rebuilt abseil package provided by fedora has different settings
than the ones if the tree is built with the sanitizer enabled. this
inconsistency leads to a crash.

to address this problem, we have to reinstate the abseil submodule, so
we can built it with the same compiler options with which we build the
tree.

in this change

* Revert "build: drop abseil submodule, replace with distribution abseil"
* update CMake building system with abseil header include settings
* bump up the abseil submodule to the latest LTS branch of abseil:
  lts_2024_01_16
* update scylla-gdb.py to adapt to the new structure of
  flat_hash_map

This reverts commit 8635d24424.

Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>

Closes scylladb/scylladb#18511

.gitmodules

@@ -6,6 +6,9 @@
 	path = swagger-ui
 	url = ../scylla-swagger-ui
 	ignore = dirty
+[submodule "abseil"]
+	path = abseil
+	url = ../abseil-cpp
 [submodule "scylla-jmx"]
 	path = tools/jmx
 	url = ../scylla-jmx

CMakeLists.txt

@@ -55,6 +55,33 @@ set(Seastar_EXCLUDE_APPS_FROM_ALL ON CACHE BOOL "" FORCE)
 set(Seastar_EXCLUDE_TESTS_FROM_ALL ON CACHE BOOL "" FORCE)
 set(Seastar_UNUSED_RESULT_ERROR ON CACHE BOOL "" FORCE)
 add_subdirectory(seastar)
+set(ABSL_PROPAGATE_CXX_STD ON CACHE BOOL "" FORCE)
+
+find_package(Sanitizers QUIET)
+set(sanitizer_cxx_flags
+    $<$<IN_LIST:$<CONFIG>,Debug;Sanitize>:$<TARGET_PROPERTY:Sanitizers::address,INTERFACE_COMPILE_OPTIONS>;$<TARGET_PROPERTY:Sanitizers::undefined_behavior,INTERFACE_COMPILE_OPTIONS>>)
+if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
+    set(ABSL_GCC_FLAGS ${sanitizer_cxx_flags})
+elseif(CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
+    set(ABSL_LLVM_FLAGS ${sanitizer_cxx_flags})
+endif()
+set(ABSL_DEFAULT_LINKOPTS
+    $<$<IN_LIST:$<CONFIG>,Debug;Sanitize>:$<TARGET_PROPERTY:Sanitizers::address,INTERFACE_LINK_LIBRARIES>;$<TARGET_PROPERTY:Sanitizers::undefined_behavior,INTERFACE_LINK_LIBRARIES>>)
+add_subdirectory(abseil)
+add_library(absl-headers INTERFACE)
+target_include_directories(absl-headers INTERFACE
+    "${PROJECT_SOURCE_DIR}/abseil")
+add_library(absl::headers ALIAS absl-headers)
+
+# Exclude absl::strerror from the default "all" target since it's not
+# used in Scylla build and, moreover, makes use of deprecated glibc APIs,
+# such as sys_nerr, which are not exposed from "stdio.h" since glibc 2.32,
+# which happens to be the case for recent Fedora distribution versions.
+#
+# Need to use the internal "absl_strerror" target name instead of namespaced
+# variant because `set_target_properties` does not understand the latter form,
+# unfortunately.
+set_target_properties(absl_strerror PROPERTIES EXCLUDE_FROM_ALL TRUE)
 
 # System libraries dependencies
 find_package(Boost REQUIRED
@@ -66,7 +93,6 @@ target_link_libraries(Boost::regex
 find_package(Lua REQUIRED)
 find_package(ZLIB REQUIRED)
 find_package(ICU COMPONENTS uc i18n REQUIRED)
-find_package(absl COMPONENTS hash raw_hash_set REQUIRED)
 find_package(fmt 9.0.0 REQUIRED)
 find_package(libdeflate REQUIRED)
 find_package(libxcrypt REQUIRED)
@@ -124,6 +150,8 @@ target_sources(scylla-main
 target_link_libraries(scylla-main
   PRIVATE
     db
+    absl::headers
+    absl::btree
     absl::hash
     absl::raw_hash_set
     Seastar::seastar
@@ -236,6 +264,7 @@ target_link_libraries(scylla PRIVATE
 
 target_link_libraries(scylla PRIVATE
     seastar
+    absl::headers
     Boost::program_options)
 
 target_include_directories(scylla PRIVATE

alternator/CMakeLists.txt

@@ -27,7 +27,8 @@ target_link_libraries(alternator
   cql3
   idl
   Seastar::seastar
-  xxHash::xxhash)
+  xxHash::xxhash
+  absl::headers)
 
 check_headers(check-headers alternator
   GLOB_RECURSE ${CMAKE_CURRENT_SOURCE_DIR}/*.hh)

api/CMakeLists.txt

@@ -72,7 +72,8 @@ target_link_libraries(api
   idl
   wasmtime_bindings
   Seastar::seastar
-  xxHash::xxhash)
+  xxHash::xxhash
+  absl::headers)
 
 check_headers(check-headers api
   GLOB_RECURSE ${CMAKE_CURRENT_SOURCE_DIR}/*.hh)

auth/CMakeLists.txt

@@ -30,6 +30,7 @@ target_link_libraries(scylla_auth
     Seastar::seastar
     xxHash::xxhash
   PRIVATE
+    absl::headers
     cql3
     idl
     wasmtime_bindings

cdc/CMakeLists.txt

@@ -11,6 +11,7 @@ target_include_directories(cdc
     ${CMAKE_SOURCE_DIR})
 target_link_libraries(cdc
   PUBLIC
+    absl::headers
     Seastar::seastar
     xxHash::xxhash
   PRIVATE

configure.py

@@ -282,7 +282,7 @@ def generate_compdb(compdb, ninja, buildfile, modes):
         # build mode-specific compdbs
         for mode in modes:
             mode_out = outdir + '/' + mode
-            submodule_compdbs = [mode_out + '/' + submodule + '/' + compdb for submodule in ['seastar']]
+            submodule_compdbs = [mode_out + '/' + submodule + '/' + compdb for submodule in ['seastar', 'abseil']]
             with open(mode_out + '/' + compdb, 'w+b') as combined_mode_specific_compdb:
                 subprocess.run(['./scripts/merge-compdb.py', outdir + '/' + mode,
                                 ninja_compdb.name] + submodule_compdbs, stdout=combined_mode_specific_compdb)
@@ -1750,6 +1750,58 @@ def configure_seastar(build_dir, mode, mode_config):
     subprocess.check_call(seastar_cmd, shell=False, cwd=cmake_dir)
 
 
+def configure_abseil(build_dir, mode, mode_config):
+    # for sanitizer cflags
+    seastar_flags = query_seastar_flags(f'{outdir}/{mode}/seastar/seastar.pc',
+                                        mode_config['build_seastar_shared_libs'],
+                                        args.staticcxx)
+    seastar_cflags = seastar_flags['seastar_cflags']
+
+    abseil_build_dir = os.path.join(build_dir, mode, 'abseil')
+
+    abseil_cflags = seastar_cflags + ' ' + modes[mode]['cxx_ld_flags']
+    cmake_mode = mode_config['cmake_build_type']
+    abseil_cmake_args = [
+        '-DCMAKE_BUILD_TYPE={}'.format(cmake_mode),
+        '-DCMAKE_INSTALL_PREFIX={}'.format(build_dir + '/inst'), # just to avoid a warning from absl
+        '-DCMAKE_C_COMPILER={}'.format(args.cc),
+        '-DCMAKE_CXX_COMPILER={}'.format(args.cxx),
+        '-DCMAKE_CXX_FLAGS_{}={}'.format(cmake_mode.upper(), abseil_cflags),
+        '-DCMAKE_EXPORT_COMPILE_COMMANDS=ON',
+        '-DCMAKE_CXX_STANDARD=20',
+        '-DABSL_PROPAGATE_CXX_STD=ON',
+    ]
+
+    abseil_cmd = ['cmake', '-G', 'Ninja', real_relpath('abseil', abseil_build_dir)] + abseil_cmake_args
+
+    os.makedirs(abseil_build_dir, exist_ok=True)
+    subprocess.check_call(abseil_cmd, shell=False, cwd=abseil_build_dir)
+
+abseil_libs = ['absl/' + lib for lib in [
+    'container/libabsl_hashtablez_sampler.a',
+    'container/libabsl_raw_hash_set.a',
+    'synchronization/libabsl_synchronization.a',
+    'synchronization/libabsl_graphcycles_internal.a',
+    'debugging/libabsl_stacktrace.a',
+    'debugging/libabsl_symbolize.a',
+    'debugging/libabsl_debugging_internal.a',
+    'debugging/libabsl_demangle_internal.a',
+    'time/libabsl_time.a',
+    'time/libabsl_time_zone.a',
+    'numeric/libabsl_int128.a',
+    'hash/libabsl_hash.a',
+    'hash/libabsl_city.a',
+    'hash/libabsl_low_level_hash.a',
+    'base/libabsl_malloc_internal.a',
+    'base/libabsl_spinlock_wait.a',
+    'base/libabsl_base.a',
+    'base/libabsl_raw_logging_internal.a',
+    'profiling/libabsl_exponential_biased.a',
+    'strings/libabsl_strings.a',
+    'strings/libabsl_strings_internal.a',
+    'base/libabsl_throw_delegate.a']]
+
+
 def query_seastar_flags(pc_file, use_shared_libs, link_static_cxx=False):
     if use_shared_libs:
         opt = '--shared'
@@ -1769,9 +1821,7 @@ def query_seastar_flags(pc_file, use_shared_libs, link_static_cxx=False):
             'seastar_testing_libs': testing_libs}
 
 pkgs = ['libsystemd',
-        'jsoncpp',
-        'absl_raw_hash_set',
-        'absl_hash']
+        'jsoncpp']
 # Lua can be provided by lua53 package on Debian-like
 # systems and by Lua on others.
 pkgs.append('lua53' if have_pkg('lua53') else 'lua')
@@ -2057,6 +2107,9 @@ def write_build_file(f,
             if has_thrift:
                 local_libs += ' ' + maybe_static(args.staticthrift, '-lthrift')
                 local_libs += ' ' + maybe_static(args.staticboost, '-lboost_system')
+            objs.extend(['$builddir/' + mode + '/' + artifact for artifact in [
+                'abseil/' + x for x in abseil_libs
+            ]])
             if binary in tests:
                 if binary in pure_boost_tests:
                     local_libs += ' ' + maybe_static(args.staticboost, '-lboost_unit_test_framework')
@@ -2265,6 +2318,12 @@ def write_build_file(f,
         f.write(f'build $builddir/{mode}/dist/tar/{scylla_product}-unified-package-{scylla_version}-{scylla_release}.tar.gz: copy $builddir/{mode}/dist/tar/{scylla_product}-unified-{scylla_version}-{scylla_release}.{arch}.tar.gz\n')
         f.write(f'build $builddir/{mode}/dist/tar/{scylla_product}-unified-{arch}-package-{scylla_version}-{scylla_release}.tar.gz: copy $builddir/{mode}/dist/tar/{scylla_product}-unified-{scylla_version}-{scylla_release}.{arch}.tar.gz\n')
 
+        for lib in abseil_libs:
+            f.write('build $builddir/{mode}/abseil/{lib}: ninja $builddir/{mode}/abseil/build.ninja\n'.format(**locals()))
+            f.write('  pool = submodule_pool\n')
+            f.write('  subdir = $builddir/{mode}/abseil\n'.format(**locals()))
+            f.write('  target = {lib}\n'.format(**locals()))
+
     checkheaders_mode = 'dev' if 'dev' in modes else modes.keys()[0]
     f.write('build checkheaders: phony || {}\n'.format(' '.join(['$builddir/{}/{}.o'.format(checkheaders_mode, hh) for hh in headers])))
 
@@ -2394,7 +2453,7 @@ def write_build_file(f,
             description = List configured modes
         build mode_list: mode_list
         default {modes_list}
-        ''').format(modes_list=' '.join(default_modes), build_ninja_list=' '.join([f'{outdir}/{mode}/{dir}/build.ninja' for mode in build_modes for dir in ['seastar']]), **globals()))
+        ''').format(modes_list=' '.join(default_modes), build_ninja_list=' '.join([f'{outdir}/{mode}/{dir}/build.ninja' for mode in build_modes for dir in ['seastar', 'abseil']]), **globals()))
     unit_test_list = set(test for test in build_artifacts if test in set(tests))
     f.write(textwrap.dedent('''\
         rule unit_test_list
@@ -2438,11 +2497,14 @@ def create_build_system(args):
         mode_config['per_src_extra_cxxflags']['release.cc'] = ' '.join(get_release_cxxflags(scylla_version, scylla_release))
 
     if not args.dist_only:
+        global user_cflags, libs
         # args.buildfile builds seastar with the rules of
         # {outdir}/{mode}/seastar/build.ninja, and
         # {outdir}/{mode}/seastar/seastar.pc is queried for building flags
         for mode, mode_config in build_modes.items():
             configure_seastar(outdir, mode, mode_config)
+            configure_abseil(outdir, mode, mode_config)
+        user_cflags += ' -isystem abseil'
 
     for mode, mode_config in build_modes.items():
         mode_config.update(query_seastar_flags(f'{outdir}/{mode}/seastar/seastar.pc',
@@ -2496,7 +2558,6 @@ def configure_using_cmake(args):
     if args.staticboost:
         settings['Boost_USE_STATIC_LIBS'] = 'ON'
 
-
     source_dir = os.path.realpath(os.path.dirname(__file__))
     build_dir = os.path.join(source_dir, 'build')
 

cql3/CMakeLists.txt

@@ -131,6 +131,7 @@ target_link_libraries(cql3
     idl
     wasmtime_bindings
     Seastar::seastar
+    absl::headers
     xxHash::xxhash
     ANTLR3::antlr3
   PRIVATE

db/CMakeLists.txt

@@ -45,6 +45,7 @@ target_link_libraries(db
     mutation
     Seastar::seastar
     xxHash::xxhash
+    absl::headers
   PRIVATE
     data_dictionary
     cql3)

gms/CMakeLists.txt

@@ -20,7 +20,8 @@ target_link_libraries(gms
     Seastar::seastar
     xxHash::xxhash
   PRIVATE
-    db)
+    db
+    absl::headers)
 
 check_headers(check-headers gms
   GLOB_RECURSE ${CMAKE_CURRENT_SOURCE_DIR}/*.hh)

idl/CMakeLists.txt

@@ -74,6 +74,9 @@ add_custom_target(idl-sources
   DEPENDS ${idl_sources})
 add_library(idl INTERFACE)
 add_dependencies(idl idl-sources)
+target_link_directories(idl
+  INTERFACE
+    absl::headers)
 target_include_directories(idl
   INTERFACE
     ${scylla_gen_build_dir})

install-dependencies.sh

@@ -48,7 +48,6 @@ debian_base_packages=(
     libunistring-dev
     libzstd-dev
     libdeflate-dev
-    libabsl-dev
     librapidxml-dev
     libcrypto++-dev
     libxxhash-dev
@@ -70,7 +69,6 @@ fedora_packages=(
     snappy-devel
     libdeflate-devel
     systemd-devel
-    abseil-cpp-devel
     cryptopp-devel
     git
     python

lang/CMakeLists.txt

@@ -16,6 +16,7 @@ target_link_libraries(lang
     Seastar::seastar
     xxHash::xxhash
   PRIVATE
+    absl::headers
     ${LUA_LIBRARIES})
 
 check_headers(check-headers lang

message/CMakeLists.txt

@@ -10,6 +10,7 @@ target_link_libraries(message
   PUBLIC
     gms
     Seastar::seastar
+    absl::headers
   PRIVATE
     idl)
 

mutation/CMakeLists.txt

@@ -21,7 +21,9 @@ target_link_libraries(mutation
   PUBLIC
     idl
     Seastar::seastar
-    xxHash::xxhash)
+    xxHash::xxhash
+  PRIVATE
+    absl::headers)
 
 check_headers(check-headers mutation
   GLOB_RECURSE ${CMAKE_CURRENT_SOURCE_DIR}/*.hh)

raft/CMakeLists.txt

@@ -12,7 +12,9 @@ target_include_directories(raft
 target_link_libraries(raft
   PUBLIC
     Seastar::seastar
-    xxHash::xxhash)
+    xxHash::xxhash
+  PRIVATE
+    absl::headers)
 
 check_headers(check-headers raft
   GLOB_RECURSE ${CMAKE_CURRENT_SOURCE_DIR}/*.hh)

repair/CMakeLists.txt

@@ -6,10 +6,12 @@ target_sources(repair
     table_check.cc)
 target_include_directories(repair
   PUBLIC
+    absl::headers
     ${CMAKE_SOURCE_DIR})
 target_link_libraries(repair
   PUBLIC
     idl
+    absl::headers
     Seastar::seastar
     xxHash::xxhash)
 

replica/CMakeLists.txt

@@ -16,6 +16,7 @@ target_include_directories(replica
 target_link_libraries(replica
   PUBLIC
     db
+    absl::headers
     wasmtime_bindings
     Seastar::seastar
     xxHash::xxhash

schema/CMakeLists.txt

@@ -12,7 +12,8 @@ target_link_libraries(schema
     cql3
     idl
     Seastar::seastar
-    xxHash::xxhash)
+    xxHash::xxhash
+    absl::headers)
 
 check_headers(check-headers schema
   GLOB_RECURSE ${CMAKE_CURRENT_SOURCE_DIR}/*.hh)

scylla-gdb.py

@@ -589,8 +589,52 @@ class flat_hash_map:
         return self.__nonzero__()
 
 
+class absl_container:
+    # absl_container is the underlying type for flat_hash_map and flat_hash_set
+    # if we need to print flat_hash_set, we should yield the element of set
+    # instead of <key, value> tuple in `__iter__()`
+    def __init__(self, ref):
+        self.val = ref
+        HasInfozShift = 1
+        self.size = ref["settings_"]["value"]["size_"] >> HasInfozShift
+
+    def __len__(self):
+        return self.size
+
+    def __iter__(self):
+        if self.size == 0:
+            return
+        capacity = int(self.val["settings_"]["value"]["capacity_"])
+        control = self.val["settings_"]["value"]["control_"]
+        # for the map the slot_type is std::pair<K, V>
+        slot_type = gdb.lookup_type(str(self.val.type.strip_typedefs()) + "::slot_type")
+        slots = self.val["settings_"]["value"]["slots_"].cast(slot_type.pointer())
+        for i in range(capacity):
+            ctrl_t = int(control[i])
+            # if the control is empty or deleted, its value is less than -1, see
+            # https://github.com/abseil/abseil-cpp/blob/c1e1b47d989978cde8c5a2a219df425b785a0c47/absl/container/internal/raw_hash_set.h#L487-L503
+            if ctrl_t == -1:
+                break
+            if ctrl_t >= 0:
+                # NOTE: this only works for flat_hash_map
+                yield slots[i]['key'], slots[i]['value']
+
+    def __nonzero__(self):
+        return self.size > 0
+
+    def __bool__(self):
+        return self.size > 0
+
+
 def unordered_map(ref):
-    return flat_hash_map(ref) if ref.type.name.startswith('flat_hash_map') else std_unordered_map(ref)
+    if ref.type.name.startswith('flat_hash_map'):
+        try:
+            return flat_hash_map(ref)
+        except gdb.error:
+            # newer absl container uses a different memory layout
+            return absl_container(ref)
+    else:
+        return std_unordered_map(ref)
 
 
 def std_priority_queue(ref):

service/CMakeLists.txt

@@ -37,6 +37,7 @@ target_include_directories(service
 target_link_libraries(service
   PUBLIC
     db
+    absl::headers
     Seastar::seastar
     xxHash::xxhash
   PRIVATE

sstables/CMakeLists.txt

@@ -31,6 +31,7 @@ target_link_libraries(sstables
   PRIVATE
     readers
     tracing
+    absl::headers
     libdeflate::libdeflate
     ZLIB::ZLIB)