[Build] Workaround for protobuf-at-head test (#35798)

- Updated `third_party/utf8_range` to 25.1 to be aligned with other protobuf (protobuf, upb, and utf8_range need to be updated at the same time as they're in the same github repo)
- Added a workaround to `extract_metadata_from_bazel_xml.py` to handle `utf8_range` migration. It used to be `@utf8_range//:utf8_range` and it will be `@com_google_protobuf//third_party/utf8_range:utf8_range` from the next Protobuf release. So the script needs this hack to support both.

Closes #35798

PiperOrigin-RevId: 603822623

third_party/utf8_range/.github/workflows/bazel_tests.yml

@@ -29,8 +29,15 @@ jobs:
 
     steps:
       - uses: actions/checkout@v2
+      - name: Setup Bazel
+        run: |
+          echo "BAZEL=bazel" >> $GITHUB_ENV
+          echo "USE_BAZEL_VERSION=5.1.1" >> $GITHUB_ENV
+      - name: Setup Bazel (macOS)
+        if: runner.os == 'macOS'
+        run: echo "BAZEL=bazelisk" >> $GITHUB_ENV
       - name: Install dependencies
         run: sudo apt update && sudo apt install -y ${{ matrix.install }}
         if: matrix.install != ''
       - name: Run tests
-        run: cd ${{ github.workspace }} && CC=${{ matrix.CC }} bazel test --test_output=errors ... ${{ matrix.flags }}
+        run: cd ${{ github.workspace }} && CC=${{ matrix.CC }} ${{ env.BAZEL }} test --test_output=errors ... ${{ matrix.flags }}

third_party/utf8_range/BUILD.bazel

@@ -4,6 +4,12 @@
 # license that can be found in the LICENSE file or at
 # https://opensource.org/licenses/MIT.
 
+DEFAULT_VISIBILITY = [
+    "//third_party/protobuf:__subpackages__",
+    "//:__subpackages__",
+    "//util/utf8/internal:__subpackages__",
+]
+
 package(default_visibility = ["//visibility:public"])
 
 licenses(["notice"])
@@ -21,6 +27,7 @@ filegroup(
     ],
     visibility = [
         "@com_google_protobuf//:__subpackages__",
+        "@upb//:__subpackages__",
     ],
 )
 

third_party/utf8_range/CMakeLists.txt

@@ -73,6 +73,11 @@ if (utf8_range_ENABLE_INSTALL)
   install(FILES "${PROJECT_BINARY_DIR}/${PROJECT_NAME}-config.cmake"
     DESTINATION "${CMAKE_INSTALL_LIBDIR}/cmake/${PROJECT_NAME}"
   )
+  configure_file(${CMAKE_CURRENT_SOURCE_DIR}/cmake/utf8_range.pc.cmake
+                 ${CMAKE_CURRENT_BINARY_DIR}/utf8_range.pc @ONLY)
+  install(
+    FILES ${CMAKE_CURRENT_BINARY_DIR}/utf8_range.pc
+    DESTINATION "${CMAKE_INSTALL_LIBDIR}/pkgconfig")
 
   # Install public headers explicitly.
   install(FILES utf8_range.h utf8_validity.h

third_party/utf8_range/cmake/utf8_range.pc.cmake

@@ -0,0 +1,11 @@
+prefix=@CMAKE_INSTALL_PREFIX@
+exec_prefix=@CMAKE_INSTALL_PREFIX@
+libdir=@CMAKE_INSTALL_FULL_LIBDIR@
+includedir=@CMAKE_INSTALL_FULL_INCLUDEDIR@
+
+Name: UTF8 Range
+Description: Google's UTF8 Library
+Version: 1.0
+Requires: absl_strings
+Libs: -L${libdir} -lutf8_validity -lutf8_range @CMAKE_THREAD_LIBS_INIT@
+Cflags: -I${includedir}

third_party/utf8_range/utf8_validity.cc

@@ -133,6 +133,7 @@ size_t ValidUTF8Span(const char* data, const char* end) {
   return err_pos + (1 - ReturnPosition);
 }
 
+#ifdef __SSE4_1__
 /* Returns the number of bytes needed to skip backwards to get to the first
    byte of codepoint.
  */
@@ -148,6 +149,7 @@ inline int CodepointSkipBackwards(int32_t codepoint_word) {
   }
   return 0;
 }
+#endif  // __SSE4_1__
 
 /* Skipping over ASCII as much as possible, per 8 bytes. It is intentional
    as most strings to check for validity consist only of 1 byte codepoints.

tools/buildgen/extract_metadata_from_bazel_xml.py

@@ -99,11 +99,13 @@ EXTERNAL_PROTO_LIBRARIES = {
 # For that we need mapping from external repo name to a corresponding
 # path to a git submodule.
 EXTERNAL_SOURCE_PREFIXES = {
-    "@utf8_range": "third_party/utf8_range",
-    "@com_googlesource_code_re2": "third_party/re2",
-    "@com_google_googletest": "third_party/googletest",
-    "@com_google_protobuf": "third_party/upb",
-    "@zlib": "third_party/zlib",
+    # TODO(veblush): Remove @utf8_range// item once protobuf is upgraded to 26.x
+    "@utf8_range//": "third_party/utf8_range",
+    "@com_googlesource_code_re2//": "third_party/re2",
+    "@com_google_googletest//": "third_party/googletest",
+    "@com_google_protobuf//upb": "third_party/upb/upb",
+    "@com_google_protobuf//third_party/utf8_range": "third_party/utf8_range",
+    "@zlib//": "third_party/zlib",
 }
 
 
@@ -209,9 +211,9 @@ def _try_extract_source_file_path(label: str) -> str:
         # This is an external source file. We are only interested in sources
         # for some of the external libraries.
         for lib_name, prefix in EXTERNAL_SOURCE_PREFIXES.items():
-            if label.startswith(lib_name + "//"):
+            if label.startswith(lib_name):
                 return (
-                    label.replace("%s//" % lib_name, prefix + "/")
+                    label.replace("%s" % lib_name, prefix)
                     .replace(":", "/")
                     .replace("//", "/")
                 )
@@ -1100,7 +1102,7 @@ _BUILD_EXTRA_METADATA = {
         "build": "all",
         "_RENAME": "upb_textformat_lib",
     },
-    "@utf8_range//:utf8_range": {
+    "@com_google_protobuf//third_party/utf8_range:utf8_range": {
         "language": "c",
         "build": "all",
         # rename to utf8_range_lib is necessary for now to avoid clash with utf8_range target in protobuf's cmake
@@ -1400,6 +1402,15 @@ tests = _exclude_unwanted_cc_tests(_extract_cc_tests(bazel_rules))
 # only very little "extra metadata" would be needed and/or it would be trivial
 # to generate it automatically.
 all_extra_metadata = {}
+# TODO(veblush): Remove this workaround once protobuf is upgraded to 26.x
+if "@com_google_protobuf//third_party/utf8_range:utf8_range" not in bazel_rules:
+    md = _BUILD_EXTRA_METADATA[
+        "@com_google_protobuf//third_party/utf8_range:utf8_range"
+    ]
+    del _BUILD_EXTRA_METADATA[
+        "@com_google_protobuf//third_party/utf8_range:utf8_range"
+    ]
+    _BUILD_EXTRA_METADATA["@utf8_range//:utf8_range"] = md
 all_extra_metadata.update(_BUILD_EXTRA_METADATA)
 all_extra_metadata.update(
     _generate_build_extra_metadata_for_tests(tests, bazel_rules)