The closing marker for doc strings is no longer required to be followed by a semicolon or newline. Additionally the closing marker may be indented, in which case the indentation will be stripped from all lines in the doc string. Array destructuring now supports reference assignments using the syntax [&$a, [$b, &$c]] = $d. The same is also supported for list. instanceof now allows literals as the first operand, in which case the result is always &false;. A new CompileError exception has been added, from which ParseError inherits. A small number of compilation errors will now throw a CompileError instead of generating a fatal error. Currently this only affects compilation errors that may be thrown by token_get_all in TOKEN_PARSE mode, but more errors may be converted in the future. Trailing commas in function and method calls are now allowed. The --with-password-argon2[=dir] configure argument now provides support for both Argon2i and Argon2id hashes in the password_hash, password_verify, password_get_info, and password_needs_rehash functions. Passwords may be hashed and verified using the PASSWORD_ARGON2ID constant. Support for both Argon2i and Argon2id in the password_* functions now requires PHP be linked against libargon2 reference library ≥ 20161029. New options have been added to customize the FPM logging: log_limit This global option can be used for setting the log limit for the logged line which allows to log messages longer than 1024 characters without wrapping. It also fixes various wrapping issues. log_buffering This global option allows an experimental logging without extra buffering. decorate_workers_output This pool option allows to disable the output decoration for workers output when catch_workers_output is enabled. bcscale can now also be used as getter to retrieve the current scale in use. Full support for LDAP Controls has been added to the LDAP querying functions and ldap_parse_result: A $serverctrls parameter to send controls to the server in ldap_add, ldap_mod_replace, ldap_mod_add, ldap_mod_del, ldap_rename, ldap_compare, ldap_delete, ldap_modify_batch, ldap_search, ldap_list and ldap_read has been added. The out parameter $serverctrls to get controls from the server in ldap_parse_result has been added. Support for LDAP_OPT_SERVER_CONTROLS and LDAP_OPT_CLIENT_CONTROLS in ldap_get_option and ldap_set_option has been fixed. Support for full case-mapping and case-folding has been added. Unlike simple case-mapping, full case-mapping may change the length of the string. For example: ]]> The different casing mapping and folding modes are available through mb_convert_case: MB_CASE_LOWER (used by mb_strtolower) MB_CASE_UPPER (used by mb_strtoupper) MB_CASE_TITLE MB_CASE_FOLD MB_CASE_LOWER_SIMPLE MB_CASE_UPPER_SIMPLE MB_CASE_TITLE_SIMPLE MB_CASE_FOLD_SIMPLE (used by case-insensitive operations) Only unconditional, language agnostic full case-mapping is performed. Case-insensitive string operations now use case-folding instead of case- mapping during comparisons. This means that more characters will be considered (case insensitively) equal now. mb_convert_case with MB_CASE_TITLE now performs title-case conversion based on the Cased and CaseIgnorable derived Unicode properties. In particular this also improves handling of quotes and apostrophes. The Multibyte String data tables have been updated for Unicode 11. The Multibyte String Functions now correctly support strings larger than 2GB. Performance of the Multibyte String extension has been significantly improved across the board. The largest improvements are in case conversion functions. The mb_ereg_* functions now support named captures. Matching functions like mb_ereg will now return named captures both using their group number and their name, similar to PCRE: \w+)', '国', $matches); // => [0 => "国", 1 => "国", "word" => "国"]; ?> ]]> Additionally, mb_ereg_replace now supports the \k<> and \k'' notations to reference named captures in the replacement string: \w+)\s*', "_\k_\k'word'_", ' foo '); // => "_foo_foo_" ?> ]]> \k<> and \k'' can also be used for numbered references, which also works with group numbers greater than 9. Support for the completion_append_character and completion_suppress_append options has been added to readline_info. These options are only available if PHP is linked against libreadline (rather than libedit).