Known Issues

Most of the changes apply to both Java and C/C++. Most changes were made under ticket #8319.

• In a MessageFormat pattern, an ASCII apostrophe can normally be written directly, rather than doubled or quadrupled. It now only begins quoted text when it precedes a character that requires quoting. A double ASCII apostrophe still yields a single one in format() output.
• Sub-messages in choice/plural/select patterns are always checked for MessageFormat syntax.
• There is a new, public parser for MessageFormat and choice/plural/select pattern strings, the class MessagePattern. It can be used in validation and translation tools, for example. The same parser is used in the message/choice/plural/select implementations.
• Both named and numbered arguments can be used in a single pattern (with named-values APIs).
• Details of the MessageFormat pattern syntax are widened (e.g., the identifier syntax), and are now documented in the class API docs.
• Nested sub-messages are parsed once, during construction or applyPattern(), rather than in each format() call.
• The MessageFormat object does not need nor create/cache Format objects for choice/plural/select arguments any more.
• The C API now supports complex arguments (choice/plural/select) with arguments in sub-messages. (Previously, nested arguments were not parsed early enough for the format() value handling.)
• The C API detects conflicting types for two arguments with the same number.

Note: The JDK-compatible apostrophe mode can be selected via MessageFormat API as well as a build-time option (see the enum ApostropheMode API doc).

• Some patterns will fail to parse, or yield different results, because of the modified ASCII apostrophe handling.
• Some patterns will fail to parse as stand-alone choice/plural/select patterns when their sub-messages do not contain correct MessageFormat syntax.
• ChoiceFormat/PluralFormat/SelectFormat do not modify selected/returned sub-message strings.
• In particular, they do not reduce ASCII apostrophes.
• This is required so that the strings can be used in a MessageFormat without having to quadruple ASCII apostrophes.
• MessageFormat.setFormat() and similar: (Setting of custom Format objects, ignoring the pattern argument specification)
• toPattern() now normally returns the original pattern string, but will throw an exception if custom formats are set
• MessageFormat.getFormat() and similar now return null for non-custom choice/plural/select arguments.
• Java Serialization of MessageFormat is incompatible between pre-ICU-4.8 and ICU-4.8-and-above versions.
• The C++ ChoiceFormat internal-array getters: getLimits(), getClosures(), getFormats() are now deprecated and disabled.

• MessageFormat API doc & User Guide
• ApostropheMode API doc
• Design doc: MessageFormat 2011q1 together with the Questions & Decisions subpage

---

Updates in ICU 4.8.1 (C and J)

• Common Locale Data Repository (CLDR) 2.0.1
• Time zone database version 2011h
• Tickets
  
• #8569 UTF-7 missing terminating shift character (C: r30175 → r30178, J: r30176 → r30177)
• #8583 Currency parsing fix / API promotion (C: r30313,r30314,r30315 → r30349, J: r30314,r30315 → r30329)
• #8589 Fallback logic used for partial location time zone format is wrong (C: r30149 → r30367, J: r30147,r30150 → r30368)
• #8624 ucol_getSortKey+strncmp and ucol_strcoll give different comparison result (C: r30220 → r30353, J: r30219,r30221 → r30330)
• #8627 Performance regression in ICU 4.8 time formatting with zones (C: r30196 → r30352, J: r30198 → r30348)
• #8637 Java assertion error in PluralFormat (C: r30259,r30260 → r30355, J: r30259 → r30288)
• #8696 Merge CLDR 2.0.1 (C: r30316,r30323,r30333 → r30341, J: r30344)
• #8709 Spanish Currency, CLDR #4038 (C: r30356,r30357 → r30361, J: r30363)
  

ICU4C Download

Changes between ICU4C 4.8 and 4.8.1:

• No API changes since ICU4J 4.8, see API Changes for changes since ICU4J 4.6
• ICU4C specific changes in 4.8.1 (See the top section for common changes)
• #5890 Darwin/MacOSX shared library fix (r30283 → r30353)
• #8265 udata races when loading full set of ICU data (ucol_open U_FILE_ACCESS_ERROR) (r30326 → r30369)
• #8535 ICU4C build failure on MinGW (r30094,r30095,r30247,r30249,r30250,r30252,r30253 → r30350)
• #8576 Add some Khmer word break tests and fix some compiler warnings (r30327,r30328 → r30369)
• #8578 Compiler warnings fixes (r30202,r30205 → r30355, r30202,r30205 → r30207)
• #8585 "pkg_gencmn fails with ""gencmn: out of memory"" on z/OS 1.11" (r30204 → r30355)
• #8593 Infinite recursion crash using UNUM_SPELLOUT to format NAN (r30258 → r30367)
• #8595 gcc 4.5.2 compile warnings about UResType (r30295,r30310 → r30352)
• #8613 fix a trivial bug found by PSV (a static analysis tool) (r30188 → r30355)
• #8621 "#include of ""unum.h"" missing ""unicode/"", breaks our build" (r30257 → r30349)
• #8622 UCONFIG_NO_SERVICE=1 doesn't work due to change in ucurr.cpp (r30256 → r30349)
• #8623 init constructors in alphaindex.cpp (r30291 → r30352)
• #8629 colldata.cpp, unsafe memcpy of UnicodeStrings (r30254 → r30355)
• #8643 runConfigureICU stomps on LDFLAGS for SolarisX86 (r30224 → r30355)
• #8645 problems building without XPLINK (r30232 → r30349)
• #8654 Spoof Check, Invisible Char check problem (r30218,r30222 → r30353)
• #8657 LESwaps.h is broken on big endian machines (r30274 → r30367)
• #8666 Regular expression assertion error (r30244 → r30349)
• #8673 Unsafe memcpy of RuleBasedBreakIterator object (r30255 → r30355)
• #8675 icu::DateFormat, icu::SimpleDateFormat fails to parse full string on second attempt in 4.8 (r30266,r30282 → r30349)
• #8681 Asymmetric search needs a bigger CE ring buffer (r30307 → r30353)
• #8687 C Warning Fixes (r30285 → r30353)
• #8691 UnicodeString::startsWith(const UChar *, -1 for NUL-terminated) fails (r30292 → r30353)