--
For tracking the implementation of tasks, task actions and nested inputs, Gradle uses the class name and an identifier for the classpath which contains the implementation.
There are some situations when Gradle is not able to track the implementation precisely:

Unknown classloader::
When the classloader which loaded the implementation has not been created by Gradle, the classpath cannot be determined.

    }

    /**
     * Adds one or more external implementation dependencies.
     *
     * @param comment A description of why the dependencies are required
     * @param dependencies The dependencies
     */
    public BuildScriptBuilder implementationDependency(@Nullable String comment, BuildInitDependency... dependencies) {
        return dependency("implementation", comment, dependencies);
    }

    /**

   * concurrency, and performance characteristics as the backing map. In essence, this factory
   * method provides a {@link Set} implementation corresponding to any {@link Map} implementation.
   * There is no need to use this method on a {@link Map} implementation that already has a
   * corresponding {@link Set} implementation (such as {@link java.util.HashMap} or {@link
   * java.util.TreeMap}).
   *

        supplierInteractions = withPerVersionStatusSupplier(metadataFile, true, "println 'Alternate implementation'")
        run 'checkDeps'

        then:
        outputContains("Providing metadata for group:projectB:2.2")
        outputContains("Providing metadata for group:projectB:1.1")
        outputContains("Alternate implementation")

    }

    @ToBeFixedForConfigurationCache

        configuration.execute(getComponents());
    }

    /**
     * This is an implementation of the {@link groovy.lang.GroovyObject}'s corresponding method.
     * The interface itself is mixed-in at runtime, but we want to keep this implementation as it
     * properly handles the dynamicLookupRoutine.
     *
     * @see AsmBackedClassGenerator.ClassBuilderImpl#addDynamicMethods

   * concurrency, and performance characteristics as the backing map. In essence, this factory
   * method provides a {@link Set} implementation corresponding to any {@link Map} implementation.
   * There is no need to use this method on a {@link Map} implementation that already has a
   * corresponding {@link Set} implementation (such as {@link java.util.HashMap} or {@link
   * java.util.TreeMap}).
   *

                    to.attribute(counted, true).attribute(artifactType, "jar")
                }
            }

            dependencies {
                implementation 'org.apache.commons:commons-lang3:3.10'
            }
        """
        file("src/main/java").createDir()
        file("src/main/java/Foo.java").createFile().text = """class Foo {}"""

- Using a `java.io.File` as the `@InputArtifact` of an artifact transform.
- Using an input with an unknown implementation.
See <<validation_problems.adoc#implementation_unknown,Cannot use an input with an unknown implementation>>.
- Missing dependencies between tasks.
See <<validation_problems.adoc#implicit_dependency,Implicit dependencies between tasks>>.

# 'enable_registration_v2' opts-in to a different implementation of op and
# kernel registration - REGISTER_OP, REGISTER_KERNEL_BUILDER, etc.
#
# This setting is currently experimental. The 'v2' implementation does _not_
# correspond to a particular, finalized design; rather, it relates to
# developing one.
#
# The current aim of the 'v2' implementation is to allow 'unused' ops and

cc_library(
    name = "variable_info",
    srcs = ["variable_info.cc"],
    hdrs = ["variable_info.h"],
    visibility = [
        ":internal",
        # We reuse VariableInfo in TFRT's implementation of TpuExecuteOp.
        "//learning/brain/tfrt/tf_tpu:__pkg__",
        "//learning/brain/tfrt/tpu_plugin:__pkg__",
        "//learning/brain/tfrt/tpu_common:__pkg__",