JDK 18 Release Notes

Starting with JDK 18, UTF-8 is the default charset for the Java SE APIs. APIs that depend on the default charset now behave consistently across all JDK implementations and independently of the user’s operating system, locale, and configuration. Specifically, java.nio.charset.Charset#defaultCharset() now returns UTF-8 charset by default. The file.encoding system property is now a part of the implementation specification, which may accept UTF-8 or COMPAT. The latter is a new property value that instructs the runtime to behave as previous releases. This change is significant to users who call APIs that depend on the default charset. Users can determine whether they'd be affected or not, by specifying -Dfile.encoding=UTF-8 as the command line option with the existing Java runtime environment.

For further details, see JEP 400

jwebserver, a command-line tool to start a minimal static web server, has been introduced. The tool and the accompanying API are located in the com.sun.net.httpserver package of the jdk.httpserver module and are designed to be used for prototyping, ad-hoc coding, and testing, particularly in educational contexts.

For further details, see JEP 408.

JEP 416 reimplements core reflection with method handles. Code that depends upon highly implementation-specific and undocumented aspects of the existing implementation might be impacted. Issues that might arise include:

• Code that inspects the internal generated reflection classes (such as, subclasses of MagicAccessorImpl) no longer works and must be updated.
• Code that attempts to break the encapsulation and change the value of the private final modifiers field of Method, Field and Constructor class to be different from the underlying member might cause a runtime error. Such code must be updated.

To mitigate this compatibility risk, you can enable the old implementation as a workaround by using -Djdk.reflect.useDirectMethodHandle=false. We will remove the old core reflection implementation in a future release. The -Djdk.reflect.useDirectMethodHandle=false workaround will stop working at that point.

For further details, see JEP 416.

Introduce a service-provider interface (SPI) for host name and address resolution, so that java.net.InetAddress can make use of resolvers other than the platform's built-in resolver. This new SPI allows replacing the operating system's native resolver, which is typically configured to use a combination of a local hosts file and the Domain Name System (DNS).

For further details, see JEP 418.

An @snippet tag for JavaDoc's Standard Doclet has been added, to simplify the inclusion of example source code in API documentation.

For further details, see JEP 413 and A Programmmer's Guide to Snippets.

Introduce an API to express vector computations that reliably compile at runtime to optimal vector instructions on supported CPU architectures, thus achieving performance superior to equivalent scalar computations.

For further details, see JEP 417.

Introduce an API by which Java programs can interoperate with code and data outside of the Java runtime. By efficiently invoking foreign functions (i.e., code outside the JVM), and by safely accessing foreign memory (i.e., memory not managed by the JVM), the API enables Java programs to call native libraries and process native data without the brittleness and danger of JNI.

For further details, see JEP 419.

Enhance the Java programming language with pattern matching for switch expressions and statements, along with extensions to the language of patterns. Extending pattern matching to switch allows an expression to be tested against a number of patterns, each with a specific action, so that complex data-oriented queries can be expressed concisely and safely.

For further details, see JEP 420.

The Z Garbage Collector now supports string deduplication (JEP 192 ).

The Serial Garbage Collector now supports string deduplication (JEP 192).

The Parallel Garbage Collector now supports string deduplication (JEP 192).

The javax.tools.JavaFileManager has been extended with two new methods, getJavaFileForOutputForOriginatingFiles and getFileForOutputForOriginatingFiles, that are used to create new files with specified originating files. The Filer uses these methods when creating new files (with Filer.createSourceFile, Filer.createClassFile, Filer.createResource) in order to pass along the files containing the originating elements.

A new method Charset.forName() that takes fallback charset object has been introduced in java.nio.charset package. The new method returns fallback if charsetName is illegal or the charset object is not available for the name. If fallback is passed as null the caller can check if the named charset was available without having to catch the exceptions thrown by the Charset.forName(name) method.

A new system property, java.properties.date, has been introduced to allow applications to control the default date comment written out by the java.util.Properties::store methods. This system property is expected to be set while launching java. Any non-empty value of this system property results in that value being used as a comment instead of the default date comment. In the absence of this system property or when the value is empty, the Properties::store methods continue to write the default date comment. An additional change has also been made in the implementation of the Properties::store methods to write out the key/value property pairs in a deterministic order. The implementation of these methods now uses the natural sort order of the property keys while writing them out. However, the implementation of these methods continues to use the iteration order when any sub-class of java.util.Properties overrides the entrySet() method to return a different Set than that returned by super.entrySet().

The combination of the deterministic ordering of the properties that get written out with the new system property is particularly useful in environments where applications require the contents of the stored properties to be reproducible. In such cases, applications are expected to provide a fixed value of their choice to this system property.

For annotation processors, the Messager interface now has methods printError, printWarning, and printNote to directly report errors, warnings, and notes, respectively.

In the javax.lang.model API, the Elements utility interface has a new method, getOutermostTypeElement, which returns the outermost class or interface syntactically enclosing an element.

The method Elements.getFileObjectOf(Element) maps from an Element to the file object used to create the element.

The SourceVersion enum class has methods to map between SourceVersion values and Runtime.Version values. In turn, Runtime.Version can be used to map from a string representation of a version to a Runtime.Version object.

Compilation Replay has been improved to make it more stable and catch up with new JVM features. Enhancements include:

• Best-effort support for hidden classes, allowing more cases involving invokedynamic, MethodHandles, and lambdas to replay successfully
• DumpReplay support for C1
• Incremental inlining support
• Numerous bug fixes to improve stability

Compilation Replay is a JVM feature of debug builds that is mainly used to reproduce crashes in the C1 or C2 JIT compilers.

JDK-8272773 introduces the VM option -XX:GCCardSizeInBytes used to set a size of the area that a card table entry covers (the "card size") that is different from the previous fixed value of 512 bytes. Permissible values are now 128, 256, and 512 bytes for all platforms, and 1024 bytes for 64 bit platforms only. The default value remains 512 bytes.

The card size impacts the amount of work to be done when searching for references into an area that is to be evacuated (for example, young generation) during garbage collection. Smaller card sizes give more precise information about the location of these references, often leading to less work during garbage collection. At the same time, however, smaller card sizes can lead to more memory usage in storing this information. The increase in memory usage might result in slower performance of maintenance work during the garbage collection.

The JDK-8275056 enhancement extends the maximum allowed heap region size from 32MB to 512MB for the G1 garbage collector. Default ergonomic heap region size selection is still limited to 32MB regions maximum. Heap region sizes beyond that must be selected manually by using the -XX:G1HeapRegionSize command line option.

This can be used to mitigate both inner and outer fragmentation issues with large objects on large heaps.

On very large heaps, using a larger heap region size may also decrease internal region management overhead and increase performance due to larger local allocation buffers.

A new JDK Flight Recorder Event, jdk.FinalizerStatistics, identifies classes at runtime that use finalizers. The event is enabled by default in the JDK (in the default.jfc and profile.jfc JFR configuration files). When enabled, JFR will emit a jdk.FinalizerStatistics event for each instantiated class with a non-empty finalize() method. The event includes: the class that overrides finalize(), that class's CodeSource, the number of times the class's finalizer has run, and the number of objects still on the heap (not yet finalized). For information about using JFR, see the User Guide.

If finalization has been disabled with the --finalization=disabled option, no jdk.FinalizerStatistics events are emitted.

PKCS#11 v3.0 adds several new APIs that support new function entry points, as well as message-based encryption for AEAD ciphers, etc. For JDK 18, the SunPKCS11 provider has been updated to support some of the new PKCS#11 v3.0 APIs. To be more specific, if the "functionList" attribute in the provider configuration file is not set, the SunPKCS11 provider will first try to locate the new PKCS#11 v3.0 C_GetInterface() method before falling back to the C_GetFunctionList() method to load the function pointers of the native PKCS#11 library. If the loaded PKCS#11 library is v3.0, then the SunPKCS11 provider will cancel crypto operations by trying the new PKCS#11 v3.0 C_SessionCancel() method instead of finishing off remaining operations and discarding the results. Support for other new PKCS#11 v3.0 APIs will be added in later releases.

New methods javax.security.auth.Subject::current and javax.security.auth.Subject::callAs have been created as replacements for existing methods javax.security.auth.Subject::getSubject and javax.security.auth.Subject::doAs. The javax.security.auth.Subject::getSubject and javax.security.auth.Subject::doAs methods are deprecated for removal because they depend on Security Manager APIs deprecated in JEP 411.

A KeyStore::getAttributes method has been added that returns the attributes of an entry without having to retrieve the entry first. This is especially useful for a private key entry that has attributes that are not protected but previously could only be retrieved with a password.

An OCSP response signed with the RSASSA-PSS algorithm is now supported.

A new -version option has been added to the keytool and jarsigner commands. This option is used to print the program version of keytool and jarsigner.

The cacerts keystore file is now a password-less PKCS #12 file. All certificates inside are not encrypted and there is no MacData for password integrity. Since the PKCS12 and JKS keystore types are interoperable, existing code that uses a JKS KeyStore to load the cacerts file with any password (including null) continue to behave as expected and can view or extract the certificates contained within the keystore.

Calling keyStore.store(outputStream, null) on a PKCS12 KeyStore creates a password-less PKCS12 file. The certificates inside the file are not encrypted and the file contains no MacData. This file can then be loaded with any password (including null) when calling keyStore.load(inputStream, password).

The SunPKCS11 provider has been enhanced to support the following crypto services and algorithms when the underlying PKCS11 library supports the corresponding PKCS#11 mechanisms:
AES/KW/NoPadding Cipher <=> CKM_AES_KEY_WRAP mechanism
AES/KW/PKCS5Padding Cipher <=> CKM_AES_KEY_WRAP_PAD mechanism
AES/KWP/NoPadding Cipher <=> CKM_AES_KEY_WRAP_KWP mechanism

The serial lint warning implemented in javac traditionally checked that a Serializable class declared a serialVersionUID field. The structural checking done by the serial lint warning has been expanded to warn for cases where declarations would cause the runtime serialization mechanism to silently ignore a mis-declared entity, rendering it ineffectual. Also, the checks include several compile-time patterns that could lead to runtime failures. The specific checks include:

• For Serializable classes and interfaces, fields and methods with names matching the special fields and methods used by the serialization mechanism are declared properly.
• Additional analagous checks are done for Externalizable types as some serialization-related methods are ineffectual there.
• A serializable class is checked that it has access to a no-arg constructor in the first non-serializable class up its superclass chain.
• For enum classes, since the serialization spec states the five serialization-related methods and two fields are all ignored, the presence of any of those in an enum class will generate a warning.
• For record classes, warnings are generated for the ineffectual declarations of serialization-related methods that serialization would ignore.
• For interfaces, if a public readObject, readObjectNoData, or writeObject method is defined, a warning is issued since a class implementing the interface will be unable to declare private versions of those methods and the methods must be private to be effective. If an interface defines default writeReplace or readResolve methods, a warning will be issued since serialization only looks up the superclass chain for those methods and not for default methods from interfaces. Since a serialPersistentFields field must be private to be effective, the presence of such a (non-private) field an in interface generates a warning.
• For serializable classes without serialPersistentFields, the type of each non-transient instance field is examined and a warning issued if the type of the field cannot be serialized. Primitive types, types declared to be serializable, and arrays can be serialized. While by the JLS all arrays are considered serializable, a warning is issued if the innermost component type is not serializable.

The new warnings can be suppressed using @SuppressWarnings("serial").

The Standard Doclet supports an --add-script option used to include a reference to an external script file in the file for each page of generated documentation.

You can now use @SuppressWarnings annotations to suppress messages from DocLint about issues in documentation comments, when it is not possible or practical to fix the issues that were found. For more details, see Suppressing Messages in the DocLint section of the javadoc Tool Guide.

The following root certificate from IdenTrust has been removed from the cacerts keystore:

+ alias name "identrustdstx3 [jdk]"

  Distinguished Name: CN=DST Root CA X3, O=Digital Signature Trust Co.

The following root certificate from Google has been removed from the cacerts keystore:

+ alias name "globalsignr2ca [jdk]"

  Distinguished Name: CN=GlobalSign, O=GlobalSign, OU=GlobalSign Root CA - R2

The java.desktop module had a few implementations of finalize() that did nothing. These methods were deprecated in Java 9 and terminally deprecated in Java 16. These methods have been removed in this release. See https://bugs.openjdk.java.net/browse/JDK-8273103 for details.

Support for pre JDK 1.4 DatagramSocketImpl implementations (DatagramSocketImpl implementations that don't support connected datagram sockets, peeking, or joining multicast groups on specific network interface) has been dropped in this release. Old implementations have not have been buildable since JDK 1.4 but implementations compiled with JDK 1.3 or older continued to be usable up to this release.

Trying to call connect or disconnect on a DatagramSocket or MulticastSocket that uses an old implementation will now throw SocketException or UncheckedIOException. Trying to call joinGroup or leaveGroup will result in an AbstractMethodError.

The system property impl.prefix has been removed. This undocumented system property dates from early JDK releases where it was possible to add an implementation of the JDK internal and non-public java.net.InetAddressImpl interface to the java.net package, and have it be used by the java.net.InetAddress API.
The InetAddressResolver SPI introduced by JEP 418 provides a standard way to implement a name and address resolver.

Legacy implementations of java.net.SocketImpl and java.net.DatagramSocketImpl have been removed from the JDK. The legacy implementation of SocketImpl has not been used by default since JDK 13, while the legacy implementation of DatagramSocketImpl has not been used by default since JDK 15. Support for system properties jdk.net.usePlainSocketImpl and jdk.net.usePlainDatagramSocketImpl used to select these implementations has also been removed. Setting these properties now has no effect.

The "default_checksum" and "safe_checksum_type" settings in the krb5.conf configuration file are no longer supported. The checksum type in a KRB_TGS_REQ message is derived from the type of the encryption key used to generate it. The "safe_checksum_type" setting was never used in Java.

The finalization mechanism has been deprecated for removal in a future release. The finalize methods in standard Java APIs, such as Object.finalize() and Enum.finalize(), have also been deprecated for removal in a future release, along with methods related to finalization, such as Runtime.runFinalization() and System.runFinalization().

Finalization remains enabled by default, but it can be disabled for testing purposes by using the command-line option --finalization=disabled introduced in this release. Maintainers of libraries and applications that rely upon finalization should migrate to other resource management techniques in their code, such as try-with-resources and cleaners.

For further details, see JEP 421.

Two javax.security.auth.Subject::doAs methods have been deprecated for removal. This is a part of the ongoing effort to remove Security Manager related APIs.

Developers that use the unsupported sun.misc.Unsafe API should be aware that the methods that return base and field offsets have been deprecated in this release. The methods objectFieldOffset, staticFieldOffset, and staticFieldBase methods are an impediment to future changes. A future release will eventually degrade or remove these methods along with the heap accessor methods.

The java.lang.invoke.VarHandle API (added in Java 9) provides a strongly typed reference to a variable that is safe and a replacement to many cases that use field offsets and the heap accessor methods. It is strongly recommended to migrate to the VarHandle API where possible. Multi-Release JARs can be used by libraries and frameworks that continue to support JDK 8 or older.

Thread.stop is terminally deprecated in this release so that it can be degraded in a future release and eventually removed. The method is inherently unsafe and has been deprecated since Java 1.2 (1998).

The options -XX:G1RSetRegionEntries and -XX:G1RSetSparseRegionEntries have been obsoleted with the changes from JDK-8017163.

JDK-8017163 implements a completely new remembered set implementation in which these two options no longer apply. In this release, neither -XX:G1RSetRegionEntries nor -XX:G1RSetSparseRegionEntries have a function, and their use will trigger an obsoletion warning.

The VM option UseBiasedLocking along with the VM options BiasedLockingStartupDelay, BiasedLockingBulkRebiasThreshold, BiasedLockingBulkRevokeThreshold, BiasedLockingDecayTime and UseOptoBiasInlining have been obsoleted. Use of these options will produce an obsolete option warning but will otherwise be ignored. Biased locking has been disabled by default and deprecated since JDK 15.

On Windows 11 and Windows Server 2022, there can be some slowness with the extraction of temporary installation files when launched from a mapped network drive. The installer will still work, but there can be a temporary delay.

The default value of the java.security.manager system property has been changed to disallow. Unless the system property is set to allow on the command line, any invocation of System.setSecurityManager(SecurityManager) with a non-null argument will throw an UnsupportedOperationException.

JARs signed with SHA-1 algorithms are now restricted by default and treated as if they were unsigned. This applies to the algorithms used to digest, sign, and optionally timestamp the JAR. It also applies to the signature and digest algorithms of the certificates in the certificate chain of the code signer and the Timestamp Authority, and any CRLs or OCSP responses that are used to verify if those certificates have been revoked.

To reduce the compatibility risk for applications that have been previously timestamped, there is one exception to this policy:

• Any JAR signed with SHA-1 algorithms and timestamped prior to January 01, 2019 will not be restricted.

This exception might be removed in a future JDK release.

Users can, at their own risk, remove these restrictions by modifying the java.security configuration file (or override it by using the java.security.properties system property) and removing "SHA1 usage SignedJAR & denyAfter 2019-01-01" from the jdk.certpath.disabledAlgorithms security property and "SHA1 denyAfter 2019-01-01" from the jdk.jar.disabledAlgorithms security property.

Prior to JDK 18, when javac compiles an inner class it always generates a private synthetic field with a name starting with this$ to hold a reference to the enclosing instance, even if the inner class does not reference its enclosing instance and the field is unused.

Starting in JDK 18, unused this$ fields are omitted; the field is only generated for inner classes that reference their enclosing instance.

For example, consider the following program:

class T {

    class I {
    }
}

Prior to JDK 18, the program would be translated as:

class T {

    class I {
        private synthetic T this$0;
        I(T this$0) {
            this.this$0 = this$0;
        }
    }
}

Starting in JDK 18, the unused this$0 field is omitted:

class T {

    class I {
        I(T this$0) {}
    }
}

Note that the form of the inner class constructor is not affected by this change.

The change may cause enclosing instances to be garbage collected sooner, if previously they were only reachable from a reference in an inner class. This is typically desirable, since it avoids a source of potential memory leaks when creating inner classes that are intended to outlive their enclosing instance.

Subclasses of java.io.Serializable are not affected by this change.

For a reference to a specific overload of an overloaded method, the javadoc tool might have linked to the wrong overload in the generated documentation. This fix resolves that issue, and the specified overload will now be used for the links in the generated documentation.

The java.io.PrintStream, PrintWriter, and OutputStreamWriter constructors that take a java.io.OutputStream and no charset now inherit the charset when the output stream is a PrintStream. This is important for usages such as:

new PrintWriter(System.out)

where it would be problematic if PrintStream didn't use the same charset as that used by System.out. This change was needed because JEP 400 makes it is possible, especially on Windows, that the encoding of System.out is not UTF-8. This would cause problems if PrintStream were wrapped with a PrintWriter that used UTF-8.

As part of this change, java.io.PrintStream now defines a charset() method to return the print stream's charset.

The java.io.ObjectInputStream.GetField.get(String name, Object val) method now throws ClassNotFoundException when the class of the object is not found. Previously, null was returned, which prevented the caller from correctly handling the case where the class was not found. The signature of GetField.get(name, val) has been updated to throw ClassNotFoundExceptionand a ClassNotFoundException exception is thrown when the class is not found.

The source compatibility risk is low. The addition of a throws ClassNotFoundException should not cause a source compatibility or a compilation warning. The GetField object and its methods are called within the context of the readObject method and include throws ClassNotFoundException.

To revert to the old behavior, a system property, jdk.serialGetFieldCnfeReturnsNull, has been added to the implementation. Setting the value to true reverts to the old behavior (returning null); leaving it unset or to any other value results in the throwing of ClassNotFoundException.

Invalid values of the command line and the security properties of jdk.serialFilter and jdk.serialFilterFactory are reported by throwing java.lang.IllegalStateException on the first use. The property values are checked when constructing java.io.ObjectInputStream or when calling the methods of java.io.ObjectInputFilter.Config including getSerialFilter() and getSerialFilterFactory(). The IllegalStateException indicates that the serial filter or serial filter factory is invalid and cannot be used; deserialization is disabled. Previously, the exception thrown was ExceptionInInitializerError.

The scope of the com.sun.jndi.ldap.object.trustSerialData system property has been extended to control the deserialization of java objects from the javaReferenceAddress LDAP attribute. This system property now controls the deserialization of java objects from the javaSerializedData and javaReferenceAddress LDAP attributes.

To prevent deserialization of java objects from these attributes, the system property can be set to false. By default, the deserialization of java objects from javaSerializedData and javaReferenceAddress attributes is allowed.

URLConnection has been fixed to return multiple header values for a given field-name in the order in which they were added.

Previously, if a URLConnection contained multiple header values for a given header field-name, when retrieved by using the HttpURLConnection::getHeaderFields and the URLConnection::getRequestProperties methods, they would be returned in the reverse order to which they were added.

This has been fixed to conform to RFC2616, which explicitly states that the order matters and thus, should be maintained.

In JDK 18, the handling of header names and values in jdk.httpserver/com.sun.net.httpserver.Headers has been reconciled. This includes the eager and consistent prohibition of null for names and values. The class represents header names and values as a key-value mapping of Map<String, List <String>>. Previously, it was possible to create a headers instance with a null key or value, which would cause undocumented exceptions when passed to the HttpServer. It was also possible to query the instance for a null key and false would be returned. With this change, all methods of the class now throw a NullPointerException if the key or value arguments are null. For more information, see https://bugs.openjdk.java.net/browse/JDK-8269296.

ReadableByteChannel::read no longer incorrectly throws ReadOnlyBufferException.

The read(ByteBuffer) method of the ReadableByteChannel returned by java.nio.channels.Channel.newChannel(InputStream) was incorrectly throwing ReadOnlyBufferException when its ByteBuffer parameter specified a read-only buffer. ReadableByteChannel::read is, however, specified in this case to throw an IllegalArgumentException. The implementation has been changed for this case to throw an IllegalArgumentException instead of a ReadOnlyBufferException, as dictated by the specification.

The ZIP file system provider has been changed to reject existing ZIP files that contain entries with "." or ".." in name elements. ZIP files with these entries cannot be used as a file system. Invoking the java.nio.file.FileSystems.newFileSystem(...) methods throw ZipException if the ZIP file contains these entries.

The IANA Time Zone Database, on which the JDK's Date/Time libraries are based, has made a tweak to some of the time zone rules in 2021c.

Note that in 2021b, which is cumulatively included in this change, some of the time zone rules prior to the year 1970 have been modified according to changes introduced with 2021b. For more details, refer to the announcement of 2021b.

JAR Index has been disabled in this release. JAR Index was an optimization to postpone downloading of JAR files when loading applets or other classes over the network. JAR Index has many long standing issues and does not interact well with newer features (such as, Multi-Release JARs and modular JARs). If a JAR file contains an INDEX.LIST file, then its contents are ignored by the application class loader and by any URLClassLoader created by user code.

The system property, jdk.net.URLClassPath.enableJarIndex, can be used re-enable the feature if required. If set to an empty string or the value "true", then JAR Index will be re-enabled. This system property is temporary. A future release will remove the JAR Index feature and the system property.

The change does not impact the jar -i option. The jar tool continues to create an index when this option is used.

DeflaterOutputStream.close() and GZIPOutputStream.finish() has been changed to close out the associated default JDK compressor before propagating a Throwable up the stack. ZIPOutputStream.closeEntry() has been changed to close out the associated default JDK compressor before propagating an IOException, not of type ZipException, up the stack.

A bug has been fixed that could cause long "Concurrent Process Non-Strong References" times with ZGC. The bug blocked the GC from making significant progress, and caused both latency and throughput issues for the Java application.

The long times could be seen in the GC logs when running with -Xlog:gc*:

[17606.140s][info][gc,phases ] GC(719) Concurrent Process Non-Strong References 25781.928ms

The JDK implementation (as supplied by the SUN provider) of X509Certificate::getSubjectAlternativeNames, X509Certificate::getIssuerAlternativeNames and X509Certificate::getExtendedKeyUsage now throws CertificateParsingException instead of returning null when the extension is non-critical and unparseable (badly encoded or contains invalid values). This change in behavior is compliant with the specification of these methods.

Support for AES/KW/NoPadding, AES/KW/PKCS5Padding and AES/KWP/NoPadding ciphers is added to SunJCE provider since jdk 17. The cipher block size for these transformations should be 8 instead of 16. In addition, for KWP mode, only the default IV, i.e. 0xA65959A6, is allowed to ensure maximum interoperability with other implementations. Other IV values will be rejected with exception during Cipher.init(...) calls.

The (D)TLS implementation in JDK now calls X509KeyManager.chooseClientAlias() only once during handshaking for client authentication, even if there are multiple algorithms requested .

Weak encryption types based on DES, 3DES, and RC4 have been removed from the default encryption types list in Kerberos. These weak encryption types can be enabled (at your own risk) by adding them to the "permitted_enctypes" property (or alternatively, the "default_tkt_enctypes" or "default_tgt_enctypes" properties) and setting "allow_weak_crypto" to "true" in the krb5.conf configuration file.

Various terms have been added to the index files in the JDK API documentation, so that these terms can be found in both the static A-Z index and in the interactive search index.

The terms are:
Language Model, Annotation Processing, Compiler API, and Compiler Tree API.

The pages generated by the Standard Doclet provide improved navigation controls when the pages are viewed on small devices.

DocLint detects and reports documentation comments that do not have a description about the associate declaration before any block tags that may be present. DocLint is a feature of the javac and javadoc tools that detects and reports issues in documentation comments.

The documentation in the Tool Guides ("man pages") for the javac and javadoc tools has been reorganized and updated. The primary documentation is now in the Tool Guide for javadoc, with information about the tool-specific command-line options being given in the Options section of the appropriate Tool Guide.

When the Standard Doclet encounters content in a documentation comment that it cannot process, it may create an element in the generated output to indicate clearly to the reader that the output at that position is not as the author intended. (This replaces the earlier behavior to show either nothing or the invalid content.) The element will contain a short summary message and may contain additional details. This is in addition to the existing behavior to report diagnostics and to return a suitable exit code.

The "Exceptions" and "Errors" tabs in documentation generated by JavaDoc have been merged into a single "Exception Classes" tab, which includes all exception classes, as defined in JLS 11.1.1.

The initialization of the file.encoding system property on non macOS platforms has been reverted to align with the behavior on or before JDK 11. This has been an issue especially on Windows where the system and user's locales are not the same.

For JVMs running in a container, OperatingSystemMXBean.getProcessCpuLoad now considers only the CPU resources available to the container when calculating CPU load. Prior to this change, the calculation included all CPUs on a host. After this change, management agents may report higher CPU usage by JVMs in containers that are constrained to a limited set of CPUs.

The default, built-in filename map in the JDK has been updated to include additional filename to MIME type mappings. Along with the additions, existing mappings of .gz and .zip have also been updated. This can result in the default, built-in java.net.FileNameMap instance returning a different MIME type from the getContentTypeFor(String fileName) method than what it returned in previous JDK releases.

java.net.URLConnection.guessContentTypeFromName(String fname), which may use this default, built-in FileNameMap instance, thus may return a different value than previously.