code can only be used as a dynamic library). On the other hand, when
using internal linking, cmd/link can generate Go binaries by itself.

In order to allow linking arbitrary object files without requiring
dynamic libraries, cgo supports an "external" linking mode too. In
external linking mode, cmd/link does not process any host object files.
Instead, it collects all the Go code and writes a single go.o object

    visibility = ["//visibility:public"],
)

# The interface library (tensorflow.dll.if.lib) for linking tensorflow DLL library (tensorflow.dll) on Windows.
# To learn more about import library (called interface library in Bazel):
#     https://docs.microsoft.com/en-us/cpp/build/linking-an-executable-to-a-dll?view=vs-2017#linking-implicitly
filegroup(
    name = "get_tensorflow_dll_import_lib",

# Link in architecture specific includes from the system; note that we cannot
# link in the whole x86_64-linux-gnu folder, as otherwise we're overlaying
# system gcc paths that we do not want to find.
# TODO(klimek): Automate linking in all non-gcc / non-kernel include
# directories.
mkdir -p "/${TARGET}/usr/include/x86_64-linux-gnu"
PYTHON_VERSIONS=("python3.9" "python3.10" "python3.11" "python3.12")
for v in "${PYTHON_VERSIONS[@]}"; do

      Class<?> cls;
      try {
        cls = classInfo.load();
      } catch (NoClassDefFoundError e) {
        // In case there were linking problems, this is probably not a class we care to test anyway.
        logger.log(Level.SEVERE, "Cannot load class " + classInfo + ", skipping...", e);
        continue;
      }
      if (!cls.isInterface()) {

      Class<?> cls;
      try {
        cls = classInfo.load();
      } catch (NoClassDefFoundError e) {
        // In case there were linking problems, this is probably not a class we care to test anyway.
        logger.log(Level.SEVERE, "Cannot load class " + classInfo + ", skipping...", e);
        continue;
      }
      if (!cls.isInterface()) {

  linking directly to the master branch. This ensures that links reference a
  specific point in time, rather than a document that may change over time.
  
  See here for guidance on getting permanent links to files: https://help.github.com/en/articles/getting-permanent-links-to-files
  
  Please use the following format for linking documentation:
  - [KEP]: <link>
  - [Usage]: <link>

  ASSERT_TRUE(o == nullptr);  // It is unsafe to move memory TF might not own.
  TF_DeleteTensor(t);
  EXPECT_TRUE(deallocator_called);
}

TEST(CAPI, LibraryLoadFunctions) {
  // TODO(b/73318067): Fix linking for the GPU test generated by the
  // tf_cuda_cc_test() bazel rule and remove the next line.
  if (!GPUDeviceName().empty()) return;

#if !defined(TENSORFLOW_NO_SHARED_OBJECTS)
  {
    // Load the library.

   *
   * - method reference: Inside Google, CacheBuilder is used from the implementation of a custom
   *   ClassLoader that is sometimes used as a system classloader. That's a problem because
   *   method-reference linking tries to look up the system classloader, and it fails because there
   *   isn't one yet.
   *
   * - lambda: Outside Google, we got a report of a similar problem in
   *   https://github.com/google/guava/issues/6565

   *
   * - method reference: Inside Google, CacheBuilder is used from the implementation of a custom
   *   ClassLoader that is sometimes used as a system classloader. That's a problem because
   *   method-reference linking tries to look up the system classloader, and it fails because there
   *   isn't one yet.
   *
   * - lambda: Outside Google, we got a report of a similar problem in
   *   https://github.com/google/guava/issues/6565

    // way for it to return a false negative would be for the target value to jump around in the map
    // such that none of the subsequent iterations observed it, despite the fact that at every point
    // in time it was present somewhere int the map. This becomes increasingly unlikely as
    // CONTAINS_VALUE_RETRIES increases, though without locking it is theoretically possible.