Introduce an API that offers statically-typed, pure-Java access to native code. This API, together with the Foreign-Memory API (JEP 393), will considerably simplify the otherwise error-prone process of binding to a native library.

For further details, see JEP 389.

Strongly encapsulate all internal elements of the JDK by default, except for critical internal APIs such as sun.misc.Unsafe. Allow end users to choose the relaxed strong encapsulation that has been the default since JDK 9.

With this change, the default value of the launcher option --illegal-access is now deny rather than permit. As a consequence, existing code that uses most internal classes, methods, or fields of the JDK will fail to run. Such code can be made to run on JDK 16 by specifying --illegal-access=permit. That option will, however, be removed in a future release.

For further details, please see JEP 396.

Introduce an API to allow Java programs to safely and efficiently access foreign memory outside of the Java heap.

For further details, see JEP 393.

Users of the value-based classes provided by the standard libraries—notably including users of the primitive wrapper classes—should avoid relying on the identity of class instances. Programmers are strongly discouraged from calling the wrapper class constructors, which are now deprecated for removal. New javac warnings discourage synchronization on value-based class instances. Runtime warnings about synchronization can also be activated, using command-line option -XX:DiagnoseSyncOnValueBasedClasses.

For further details, see JEP 390.

A new method, invokeDefault, has been added to the java.lang.reflect.InvocationHandler interface to allow a default method defined in a proxy interface to be invoked.

Provides support for Unix domain sockets (AF_UNIX) in the java.nio.channels, SocketChannel, and ServerSocketChannel classes.

See JEP-380 for more information.

See also the following release note for information about the limitations in the support on Windows in JDK16.

A new formatter pattern, letter 'B', and its supporting method have been added to java.time.format.DateTimeFormatter/DateTimeFormatterBuilder classes. The pattern and method translate day periods defined in Unicode Consortium's CLDR (https://unicode.org/reports/tr35/tr35-dates.html#dayPeriods). Applications can now express periods in a day, such as "in the morning" or "at night", not just am/pm. The following example demonstrates translating the day periods:

DateTimeFormatter.ofPattern("B").format(LocalTime.now())

This example produces day period text depending on the time of the day and locale.

A new method toList has been added to the java.util.Stream interface. This introduces a potential source incompatibility with classes that implement or interfaces that extend the Stream interface and that also statically import a toList method from elsewhere, for example, Collectors.toList. References to such methods must be changed to use a qualified name instead of a static import.

Provides an initial iteration of an incubator module, jdk.incubator.vector, to express vector computations that reliably compile at runtime to optimal vector hardware instructions on supported CPU architectures and thus achieve superior performance to equivalent scalar computations.

For further details, see JEP 338.

The CompileCommand flag has an option type that has been used for a collection of sub commands. These commands weren't verified for validity so spelling mistakes lead to the command being ignored. They had the form:

-XX:CompileCommand=option,<method pattern>,<option name>,<value type>,<value>

All option commands now exist as ordinary commands with this form:

-XX:CompileCommand=<option name>,<method pattern>,<value>

The option name is verified and the type is inferred. Helpful error messages are given if the command name doesn't exist, or if the value doesn't match the type of the command. All command names are case insensitive.

The old syntax for option commands can still be used. Verification that the option name, value type, and value is consistent has been added.

All available options can be listed with:

-XX:CompileCommand=help

The Z Garbage Collector now processes thread stacks concurrently. This allows all roots in the JVM to be processed by ZGC in a concurrent phase instead of stop-the-world pauses. The amount of work done in ZGC pauses has now become constant and typically not exceeding a few hundred microseconds.

For further details, see JEP 376.

This new feature is always enabled and changes the time when G1 returns Java heap memory to the operating system. G1 still makes sizing decisions during the GC pause, but offloads the expensive work to a thread running concurrently with the Java application.

A new JFR event, jdk.ObjectAllocationSample, is introduced to allow always-on, low-overhead allocation profiling. The event is enabled in both the default and profile configurations, with a maximum rate of 150 and 300 events/s, respectively. Events jdk.ObjectAllocationInNewTLAB and jdk.ObjectAllocationOutsideTLAB also detail memory allocations, but have comparatively higher overhead. They were previously disabled in the default configuration but are now also disabled in the profile configuration to reduce the overhead further.

The impact of this change can be seen in JDK Mission Control (JMC) 8.0, or earlier releases, where the 'TLAB Allocations' page will be missing data. The recommendation is to upgrade to a later version of JMC when the jdk.ObjectAllocationSample support is available. Meanwhile, it is possible to enable the jdk.ObjectAllocationInNewTLAB and jdk.ObjectAllocationOutsideTLAB events in the JMC recording wizard, or edit the .jfc file manually.

JEP 387 "Elastic Metaspace" overhauls the VM-internal metaspace- and class-space-implementation. Less memory is used for class metadata. The savings effect is mostly noticeable in scenarios involving lots of small grained class loaders. Upon class unloading, memory is timely returned to the operating system.

A switch is added to fine-tune metaspace reclamation: -XX:MetaspaceReclaimPolicy=(balanced|aggressive|none). balanced, the default, causes the VM to reclaim memory while keeping computational overhead minimal; aggressive moderately increases the reclaim rate at the cost of somewhat more expensive bookkeeping; none switches reclamation off altogether.

The switches InitialBootClassLoaderMetaspaceSize and UseLargePagesInMetaspace have been deprecated.

This enhancement includes two main changes:

1. The JarSigner API and the jarsigner tool now support signing a JAR file with an RSASSA-PSS or EdDSA key.

2. Instead of signing the .SF file directly, jarsigner creates a SignerInfo signedAttributes field which contains ContentType, MessageDigest, SigningTime, and CMSAlgorithmProtection. The field will not be generated if an alternative signing mechanism is specified by the jarsigner -altsigner option. Please note that although this field was not generated by jarsigner before this code change, it has always been supported when parsing the signature. This means newly signed JAR files with the field can be verified by earlier JDK releases.

SUN, SunRsaSign, and SunEC provider have been enhanced to support SHA-3 based signature algorithms. DSA signatures, RSA, and ECDSA signature implementations with SHA-3 family of digests are now supported through these providers. In addition, RSASSA-PSS signature implementation from SunRsaSign provider can recognize SHA-3 family of digests when specified in signature parameters.

When signing a file that contains POSIX file permission or symlink attributes, jarsigner now preserves these attributes in the newly signed file but warns that these attributes are unsigned and not protected by the signature. The same warning is printed during the jarsigner -verify operation for such files.

Note that the jar tool does not read/write these attributes. This change is more visible to tools like unzip where these attributes are preserved.

The -trustcacerts and -keystore options have been added to the -printcert and -printcrl commands of the keytool utility. The -printcert command does not check for the weakness of a certificate's signature algorithm if it is a trusted certificate in the user's keystore or in the cacerts keystore. The -printcrl command verifies the CRL using a certificate from the user's keystore or the cacerts keystore, and will print out a warning if it cannot be verified.

The SunPKCS11 provider has been updated with SHA-3 algorithm support. Additional KeyGenerator support for Hmac using message digests other than SHA-3 has also been added. When the corresponding PKCS11 mechanisms are supported by the underlying PKCS11 library, the SunPKCS11 provider now supports the following additional algorithms:

• MessageDigest: SHA3-224, SHA3-256, SHA3-384, SHA3-512
• Mac: HmacSHA3-224, HmacSHA3-256, HmacSHA3-384, HmacSHA3-512
• Signature: SHA3-224withDSA, SHA3-256withDSA, SHA3-384withDSA, SHA3-512withDSA, SHA3-224withDSAinP1363Format, SHA3-256withDSAinP1363Format, SHA3-384withDSAinP1363Format, SHA3-512withDSAinP1363Format, SHA3-224withECDSA, SHA3-256withECDSA, SHA3-384withECDSA, SHA3-512withECDSA, SHA3-224withECDSAinP1363Format, SHA3-256withECDSAinP1363Format, SHA3-384withECDSAinP1363Format, SHA3-512withECDSAinP1363Format, SHA3-224withRSA, SHA3-256withRSA, SHA3-384withRSA, SHA3-512withRSA, SHA3-224withRSASSA-PSS, SHA3-256withRSASSA-PSS, SHA3-384withRSASSA-PSS, SHA3-512withRSASSA-PSS.
• KeyGenerator: HmacMD5, HmacSHA1, HmacSHA224, HmacSHA256, HmacSHA384, HmacSHA512, HmacSHA512/224, HmacSHA512/256, HmacSHA3-224, HmacSHA3-256, HmacSHA3-384, HmacSHA3-512.

A new system property, jdk.tls.maxHandshakeMessageSize, has been added to set the maximum allowed size for the handshake message in TLS/DTLS handshaking. The default value of the system property is 32768 (32 kilobytes).

A new system property, jdk.tls.maxCertificateChainLength, has been added to set the maximum allowed length of the certificate chain in TLS/DTLS handshaking. The default value of the system property is 10.

Certain TLS ALPN values couldn't be properly read or written by the SunJSSE provider. This is due to the choice of Strings as the API interface and the undocumented internal use of the UTF-8 Character Set which converts characters larger than U+00007F (7-bit ASCII) into multi-byte arrays that may not be expected by a peer.

ALPN values are now represented using the network byte representation expected by the peer, which should require no modification for standard 7-bit ASCII-based character Strings. However, SunJSSE now encodes/decodes String characters as 8-bit ISO_8859_1/LATIN-1 characters. This means applications that used characters above U+000007F that were previously encoded using UTF-8 may need to either be modified to perform the UTF-8 conversion, or set the Java security property jdk.tls.alpnCharset to "UTF-8" revert the behavior.

The SunJSSE provider now supports the use of the EdDSA signature algorithm. Specifically SunJSSE may use certificates containing EdDSA keys for server side and client side authentication and may use certificates signed with the EdDSA algorithm. Additionally, EdDSA signatures are supported for TLS handshake messages that require digital signatures.

Sealed classes and interfaces have been previewed again in JDK 16, initially added to the Java language in JDK 15. Sealed classes and interfaces restrict which other classes or interfaces may extend or implement them.

For further details, see JEP 397.

Records have been added to the Java language. Records are a new kind of class in the Java language. They act as transparent carriers for immutable data with less ceremony than normal classes.

Since nested classes were first introduced to Java, with the exception of static final fields initialized by constant expressions, nested class declarations that are inner have been prohibited from declaring static members. This restriction applies to non-static member classes, local classes, and anonymous classes.

JEP 384: Records (Second Preview) added support for local interfaces, enum classes, and record classes, all of which are static definitions. This was a well-received enhancement, permitting coding styles that reduce the scope of certain declarations to local contexts.

While JEP 384 allowed for static local classes and interfaces, it did not relax the restriction on static member classes and interfaces of inner classes. An inner class could declare a static interface inside one of its method bodies, but not as a class member.

As a natural next step, JEP 395 further relaxes nesting restrictions, and permits static classes, methods, fields, etc., to be declared within inner classes.

For further details, see JEP 395.

Pattern matching for the instanceof operator has been made a final and permanent feature of the Java language in JDK 16. Pattern matching allows common logic in a Java program to be expressed more concisely and safely, namely the conditional extraction of components from objects.

For further details, see JEP 394.

Provides the jpackage tool, for packaging self-contained Java applications. The jpackage tool was introduced as an incubating tool in JDK 14 by JEP 343. It remained an incubating tool in JDK 15, to allow time for additional feedback. It has been promoted in JDK 16 from incubation to a production-ready feature. As a consequence of this transition, the name of the jpackage module has changed from jdk.incubator.jpackage to jdk.jpackage.

For further details, see JEP 392.

The non-public class java.awt.PeerFixer has been removed in this release. This class was used to provide deserialization support of ScrollPane objects created prior JDK 1.1.1.

The Java Ahead-of-Time compilation experimental tool jaotc has been removed. Using HotSpot VM options defined by JEP295 produce a not supported option warning but will otherwise be ignored.

The experimental Java-based JIT compiler, Graal JEP317, has been removed. Attempting to use it produces a JVMCI error: JVMCI compiler 'graal' not found.

When Unified Logging was added in Java 9, a number of tracing flags were deprecated and mapped to their unified logging equivalent. These flags are now obsolete and will no longer be converted automatically to enable unified logging. To continue getting the same logging output, you must explicitly replace the use of these flags with their unified logging equivalent.

The following root certificates with weak 1024-bit RSA public keys have been removed from the cacerts keystore:

+ alias name "thawtepremiumserverca [jdk]"

  Distinguished Name: EMAILADDRESS=premium-server@thawte.com, 
CN=Thawte Premium Server CA, OU=Certification Services Division, 
O=Thawte Consulting cc, L=Cape Town, ST=Western Cape, C=ZA

+ alias name "verisignclass2g2ca [jdk]"
  Distinguished Name: OU=VeriSign Trust Network, OU="(c) 1998 VeriSign, Inc. - For authorized use only", 
OU=Class 2 Public Primary Certification Authority - G2, 
O="VeriSign, Inc.", C=US

+ alias name "verisignclass3ca [jdk]"
  Distinguished Name: OU=Class 3 Public Primary Certification Authority, O="VeriSign, Inc.", C=US

+ alias name "verisignclass3g2ca [jdk]"
  Distinguished Name: OU=VeriSign Trust Network, OU="(c) 1998 VeriSign, Inc. - For authorized use only", 
OU=Class 3 Public Primary Certification Authority - G2, O="VeriSign, Inc.", C=US

+ alias name "verisigntsaca [jdk]"
  Distinguished Name: CN=Thawte Timestamping CA, OU=Thawte Certification, O=Thawte, L=Durbanville, ST=Western Cape, C=ZA

The SunEC provider no longer supports the following elliptic curves that are either obsolete or not implemented using modern formulas and techniques:

secp112r1, secp112r2, secp128r1, secp128r2, secp160k1, secp160r1, secp160r2,

secp192k1, secp192r1, secp224k1, secp224r1, secp256k1, sect113r1, sect113r2, sect131r1, 
sect131r2, sect163k1, sect163r1, sect163r2, sect193r1, sect193r2, sect233k1, sect233r1, sect239k1,
sect283k1, sect283r1, sect409k1, sect409r1, sect571k1, sect571r1, X9.62 c2tnb191v1,
X9.62 c2tnb191v2, X9.62 c2tnb191v3, X9.62 c2tnb239v1, X9.62 c2tnb239v2, X9.62 c2tnb239v3, 
X9.62 c2tnb359v1, X9.62 c2tnb431r1, X9.62 prime192v2, X9.62 prime192v3, X9.62 prime239v1, 
X9.62 prime239v2, X9.62 prime239v3, brainpoolP256r1 brainpoolP320r1, brainpoolP384r1,
brainpoolP512r1

To continue using any of these curves, users must find third-party alternatives.

The stop, destroy, isDestroyed, setDaemon and isDaemon methods defined by java.lang.ThreadGroup have been terminally deprecated in this release.

ThreadGroup::stop is inherently unsafe and has been deprecated since Java 1.2. The method is now terminally deprecated and will be removed in a future release.

The API and mechanism for destroying a ThreadGroup is inherently flawed. The methods that support explicitly or automatically destroying a thread group have been terminally deprecated and will be removed in a future release.

The signal-chaining facility was introduced in JDK 1.4 and supported three different Linux signal-handling API's: sigset, signal and sigaction. Only sigaction is a cross-platform, supported, API for multi-threaded processes. Both signal and sigset are considered obsolete on those platforms that still define them. Consequently, the use of signal and sigset with the signal-chaining facility are now deprecated, and support for their use will be removed in a future release.

The following APIs have been deprecated:

java.security.cert.X509Certificate.getIssuerDN()

java.security.cert.X509Certificate.getSubjectDN()
java.security.cert.X509CRL.getIssuerDN()
java.security.cert.X509CertSelector.setIssuer(String)
java.security.cert.X509CertSelector.setSubject(String)
java.security.cert.X509CertSelector.getIssuerAsString()
java.security.cert.X509CertSelector.getSubjectAsString()
java.security.cert.X509CRLSelector.addIssuerName(String)

These APIs either take or return Distinguished Names as Principal or String objects and can cause issues due to loss of encoding information or differences when comparing names across different Principal implementations. All of them have alternative APIs that use X500Principal objects instead.

The definition of line terminator has been extended to include end of stream, or one of the previously defined line terminators '\n', '\r', or '\r\n' followed immediately by end of stream. This changes behavior in certain cases. For example, a file containing the lines

line 1\n

line 2\n
line 3

would, before this change, have been considered to contain two lines, each terminated by '\n'. After this change, it is considered to contain three lines, the third being terminated by end of stream.

The deserialization of java.lang.reflect.Proxy objects can be limited by setting the system property jdk.serialProxyInterfaceLimit. The limit is the maximum number of interfaces allowed per Proxy in the stream. Setting the limit to zero prevents any Proxies from being deserialized including Annotations, a limit of less than 2 might interfere with RMI operations.

Case insensitive operations in java.lang.String class now correctly do case insensitive comparisons for supplementary characters (characters which have code point values over U+FFFF). For details, see the updates to the methods:

   - compareToIgnoreCase(String other)

   - equalsIgnoreCase(String other)
   - regionMatches(boolean ignoreCase, ...)

For example,

"\ud801\udc00".equalsIgnoreCase("\ud801\udc28")

returns true, because '𐐀' ("\ud801\udc00") and '𐐨' ("\ud801\udc28") are equal to each other character in case insensitive comparison.

Module::getPackages returns the set of package names for the packages in this module.

For unnamed modules, the specification and implementation have been corrected to return the names of the packages defined in the unnamed module. Prior to Java SE 16, Module::getPackages returned the set of package names for the packages defined in the module's class loader if the module was an unnamed module and the result included packages of other named modules defined in the module's class loader, if any.

All proxy classes are not open for reflective access in Java SE 16.

Prior to Java SE 16, if java.lang.reflect.Proxy was used to implement only public exported proxy interfaces, the proxy class was generated in an unnamed module which was open for reflective access. In Java SE 16, the proxy class is generated in an exported package in a named module. Programs that assume the private members of a proxy class can be made accessible via setAccessible(true) will fail with InaccessibleObjectException. Proxy classes are already defined in dynamic modules in other cases since Java SE 9. Such programs would already fail when it calls setAccessible(true) to those proxy classes prior to this change.

In this release, the default HttpClient returns cancelable futures.

The default HttpClient is created by a call to HttpClient.newHttpClient(), or by invoking the build method on builders returned by HttpClient.newBuilder(). The implementation of the sendAsync methods in the default HttpClient now return CompletableFuture objects that are cancelable. Invoking cancel(true) on a cancelable future that is not completed, attempts to cancel the HTTP exchange in an effort to release underlying resources as soon as possible. More information can be obtained by reading the API documentation of the HttpClient::sendAsync methods.

Creation of an instance of java.net.http.HttpClient may fail with UncheckedIOException if the underlying resources required by the implementation cannot be allocated.

Typically, this may happen if the underlying resources required to opening a java.nio.channels.Selector are not available. In this case Selector.open() will throw an IOException which the default implementation of java.net.http.HttpClient will wrap in an UncheckedIOException. The API documentation of HttpClient.newHttpClient() and HttpClient.newBuilder().build() have been updated to specify that UncheckedIOException may be thrown if the underlying resources required by the implementation cannot be allocated. Prior to this change, an undocumented InternalError would have been thrown.

The behavior of the HttpPrincipal::getName method has been updated to return the name in the correct format as outlined in its specification, i.e. "realm:username". Previously, the method incorrectly returned "username" only.

JEP 380: Unix-Domain Socket Channels has been integrated into JDK 16, but due to a problem with the underlying implementation in Windows 2019 Server, it has been disabled on that platform in this release. The functionality is supported on Linux, macOS and on Windows operating systems whose build number is 18362 or greater. This includes Windows 10 (Version 1903) and newer. There are no known issues with Windows 10 (Version 1903). However, some newer versions of Windows 10 may also show the problem, which affects the ability of some native tools (such as cygwin) to access and delete socket files. Other Java APIs are also affected by the problem, but most native Windows tools are not affected by it. More information can be seen in bugs JDK-8259014 and JDK-8252971. This issue will be resolved in JDK 17 and in a future update to JDK 16.

The var args form of Path.of() and Paths.get() method are changed in this release to throw NullPointerException consistently when the first parameter is null. Historically these methods missed a null check on the first parameter when invoked with more than one parameter.

DataInputStream.readFully(byte[] b, int off, int len) now correctly throws specified NullPointerException and IndexOutOfBoundsException.

DataInputStream.readFully(byte[] b, int off, int len) is specified to throw a NullPointerException if b is null, and an IndexOutOfBoundsException if off is negative, len is negative, or len is greater than b.length - off. For example if b was null and len was zero, then the method would fail silently without throwing a NullPointerException. Also if b was non-null, off was negative and len was zero, then the method would fail silently without throwing an IndexOutOfBoundsException. The implementation has been corrected to throw the respective exceptions as specified.

Following the JDK's update to tzdata2020b, the long-obsolete files named pacificnew and systemv have been removed. As a result, the "US/Pacific-New" Zone name declared in the pacificnew data file is no longer available for use.

Information regarding this update can be viewed at https://mm.icann.org/pipermail/tz-announce/2020-October/000059.html

Format string specifiers now throw exceptions when given values outside of valid ranges of values.

The Formatter class in java.util defines format specifiers such as argument indexes, argument widths, and argument precisions. Numeric values that are invalid (zero for argument index) or too large (beyond the size of an int) could create unexpected results with undefined behavior. This update gives explicit value ranges for these format specifiers. Widths and indexes will be valid from [1, Integer.MAX_VALUE] and precision fields will be valid from [0, Integer.MAX_VALUE]. Values outside of these ranges will result in IllegalFormatException or one of its subclasses being thrown. Note that argument indexes of zero prior to this change did not throw an exception and produced behavior that was undefined, but had the appearance of correctness despite causing side effects that were not obvious.

Prior to JDK 16, ZipOutputStream.putnextEntry() would not recalculate the compressed size for a compressed (DEFLATED) entry. This could result in the ZipException, "invalid entry compressed size", being thrown if the current ZLIB implementation being used when ZipOutputStream.putNextEntry() was called differed from the implementation at the time when the entry was added to the original ZIP file.

Starting with JDK 16, if the compressed size has not been explicitly set with the ZipEntry.setCompressedSize(long) method when writing a compressed (DEFLATED) entry, then the compressed size is set to the actual compressed size after deflation.

Prior to JDK 16, GZIPOutputStream set the OS field in the GZIP header to 0 (meaning FAT filesystem), which does not match the default value specified in section 2.3.1.2 of the GZIP file format specification version 4.3 (RFC 1952).

As of JDK 16, the GZIP OS Header Field is set to 255, which is the default value as defined in RFC 1952.

In this release, java.util.logging.LogRecord has been updated to support long thread ids.

LogRecord::getThreadID and LogRecord::setThreadID are deprecated. New accessors, LogRecord::getLongThreadID and LogRecord::setLongThreadID, are provided and should be used instead.

The serial field threadID has been deprecated and a new serial field longThreadID that can support long values has been introduced. The deprecated threadID field is kept in the serial form for backward compatibility. Long thread ids that are less than Integer.MAX_VALUE are directly mapped to threadID and longThreadID has the same value. For longThreadID greater than Integer.MAX_VALUE, a new negative value for the int threadID is synthesized by using a deterministic algorithm based on the longThreadID hash. The resulting value is guaranteed to be a negative value.

Enhancement JDK-8176894 inadvertently introduced erroneous behavior in the TreeMap.computeIfAbsent method. The other TreeMap methods that were modified by this enhancement are unaffected. The erroneous behavior is that, if the map contains an existing mapping whose value is null, the computeIfAbsent method immediately returns null. To conform with the specification, computeIfAbsent should instead call the mapping function and update the map with the function's result.

Locale data based on Unicode Consortium's CLDR has been upgraded to their version 38. For the detailed locale data changes, please refer to the Unicode Consortium's CLDR release notes:

• http://cldr.unicode.org/index/downloads/cldr-38

A new environment property, jdk.jndi.ldap.mechsAllowedToSendCredentials, has been added to control which LDAP authentication mechanisms are allowed to send credentials over clear LDAP connections - a connection not secured with TLS. An encrypted LDAP connection is a connection opened by using ldaps scheme, or a connection opened by using ldap scheme and then upgraded to TLS with a STARTTLS extended operation.

The value of the property, which is by default not set, is a comma separated list of the mechanism names that are permitted to authenticate over a clear connection. If a value is not specified for the property, then all mechanisms are allowed. If the specified value is an empty list, then no mechanisms are allowed (except for none and anonymous). The default value for this property is 'null' ( i.e. System.getProperty("jdk.jndi.ldap.mechsAllowedToSendCredentials") returns 'null'). To explicitly permit all mechanisms to authenticate over a clear connection, the property value can be set to "all". If a connection is downgraded from encrypted to clear, then only the mechanisms that are explicitly permitted are allowed.

The property can be supplied to the LDAP context environment map, or set globally as a system property. When both are supplied, the environment map takes precedence.

Note: none and anonymous authentication mechanisms are exempted from these rules and are always allowed regardless of the property value.

A new JNDI environment property “com.sun.jndi.ldap.tls.cbtype” has been added to enable TLS Channel Binding data in LDAP authentication over SSL/TLS protocol to the Windows AD server. Possible value is “tls-server-end-point” - Channel Binding data is created on the base of the TLS server certificate. See the module description of the java.naming module.

The Shenandoah GC now processes all soft, weak and phantom references as well as finalizers concurrently with the running Java application, instead of at a GC pause. This reduces GC induced latency, especially with workloads that churn many such references or finalizers.

For improved performance, JVM/TI ObjectFree events are no longer posted within GC pauses. The events are still posted as requested, and will be posted before ObjectFree events are enabled or disabled with SetNotificationMode. SetNotificationMode can be used to explicitly flush ObjectFree events, if needed.

As a result of changes to section 5.3.5 of the Java Virtual Machine Specification, a java.lang.IncompatibleClassChangeError exception is thrown instead of a java.lang.VerifyError exception when defining a class whose super class is final and when defining a class that tries to override a final method.

When synchronization is performed on an object, an association is established between the object and the object monitor that implements the synchronization. In the past, the reference from a monitor to its associated object was a strong reference. These strong references would be observable through JVM TI functions that walk the heap (reported as JVMTI_HEAP_ROOT_MONITOR or JVMTI_HEAP_REFERENCE_MONITOR) and in heap dumps (reported as HPROF_GC_ROOT_MONITOR_USED). As of this release, a weak reference is used. These are not observable to JVM TI or heap dumps. Consequently, JVMTI_HEAP_ROOT_MONITOR, JVMTI_HEAP_REFERENCE_MONITOR and HPROF_GC_ROOT_MONITOR_USED are longer reported.

The following root certificate has been added to the cacerts truststore:

+ Entrust

  + entrustrootcag4
    DN: CN=Entrust Root Certification Authority - G4, OU="(c) 2015 Entrust, Inc. - for authorized use only", 
           OU=See www.entrust.net/legal-terms, O="Entrust, Inc.", C=US

The following root certificates have been added to the cacerts truststore:

+ SSL Corporation

  + sslrootrsaca
    DN: CN=SSL.com Root Certification Authority RSA, O=SSL Corporation, L=Houston, ST=Texas, C=US

  + sslrootevrsaca
    DN: CN=SSL.com EV Root Certification Authority RSA R2, O=SSL Corporation, L=Houston, ST=Texas, C=US

  + sslrooteccca
    DN: CN=SSL.com Root Certification Authority ECC, O=SSL Corporation, L=Houston, ST=Texas, C=US

The default encryption and MAC algorithms used in a PKCS #12 keystore have been updated. The new algorithms are based on AES-256 and SHA-256 and are stronger than the old algorithms that were based on RC2, DESede, and SHA-1. See the security properties starting with keystore.pkcs12 in the java.security file for detailed information.

For compatibility, a new system property named keystore.pkcs12.legacy is defined that will revert the algorithms to use the older, weaker algorithms. There is no value defined for this property.

TLS 1.0 and 1.1 are versions of the TLS protocol that are no longer considered secure and have been superseded by more secure and modern versions (TLS 1.2 and 1.3).

These versions have now been disabled by default. If you encounter issues, you can, at your own risk, re-enable the versions by removing "TLSv1" and/or "TLSv1.1" from the jdk.tls.disabledAlgorithms security property in the java.security configuration file.

Prior to JDK 16, the javac compiler accepted annotations declared as local interfaces. Now the JLS 16 at section 14.3 forbids it. For example, the javac compiler had accepted code such as:

    class C {

        void m() {
            @interface A {}
        }
    }

This code is no longer acceptable according to Section [14.3] of the JLS 16: "A local interface may be a normal interface (§9.1), but not an annotation interface (§9.6)."

To fix this issue, the javac compiler implementation has been synchronized with the JLS 16, so that annotation interfaces cannot be declared as local interfaces.

Prior to JDK 16, the javac compiler accepted C-style array declarations in record components. The JLS 16 now forbids it. In particular, the compiler had accepted code such as:

    record R(int i[]) {}

This code is no longer acceptable according to the specification for JDK 16. Section 8.10.1 of the JLS 16 defines the syntax of a record component as:

RecordComponent:

    {RecordComponentModifier} UnannType Identifier 
    VariableArityRecordComponent 
VariableArityRecordComponent:
    {RecordComponentModifier} UnannType { Annotation } ... Identifier
RecordComponentModifier:
    Annotation

which clearly forbids C-style array declarations in record components. To fix this issue, the compiler has been synchronized with the JLS 16, so that C-style array declarations are no longer allowed in record components.

The code that provides the DocLint feature in javac has been moved from the jdk.compiler module to the jdk.javadoc module. For most users, there will be no impact. However, any user wanting to build a minimal image for running javac with DocLint enabled will need to ensure that the jdk.javadoc module is available when linking the image.

The interactive search feature in API documentation generated by the JavaDoc Standard Doclet has been improved to find additional matches that differ only in case. This is useful when the exact spelling or capitalization of a name is not known.

For full details on JavaDoc Search, see the link to the JavaDoc Search Specification that is found on at the bottom of the HELP page in any API documentation.

API documentation generated by the JavaDoc Standard Doclet will now automatically link to the appropriate set of platform classes, based on the value of the --source or --release option.

The behavior can be disabled by using the new --no-platform-links option.

The URLs for the different versions of the platform documentation can be configured by the new --link-platform-properties option.

API documentation generated by the JavaDoc Standard Doclet has been updated so that it can be viewed more easily on small devices.

For many simple methods, the first sentence is often a near duplicate of the text provided for the @return tag. To help avoid such duplication, the @return tag has been enhanced so that it can be used as an inline tag at the beginning of the comment, to provide the text for both the first sentence, and the description of the return value.

In line with guidelines described elsewhere 1, the Standard Doclet uses the following terms when generating documentation for JDK 16 or later. For consistency, the older terms will continue to be used when generating documentation for older releases.

• Classes and Interfaces (instead of Types)
• Annotation Interface (instead of Annotation Type)
• Enum Class (instead of Enum)
• Record Class (instead of Record)

On Linux, process memory information has been added to both JVM crash reports (hs_err files) and the VM.info diagnostic of jcmd. This information contains the process' virtual size, its resident set size, and how much memory was swapped out. If the JVM uses glibc, the size of glibc outstanding allocations and retained memory are printed, as well as the glibc tunables.