Update ICU to 78 - Part 2: Microsoft patch/code changes

icu-patches/patches/021-MSFT-Patch_ICU_Cldr2Icu_remove_icu4j_directory_validation.patch

@@ -0,0 +1,20 @@
+From 0000000000000000000000000000000000000000 Mon Sep 17 00:00:00 2001
+From: Arvind Oruganti <aoruganti@microsoft.com>
+Date: Thu, 14 May 2026 14:30:00 +0530
+Subject: MSFT-PATCH: Remove icu4j directory validation from Cldr2Icu (icu4c-only fork)
+
+diff --git a/icu/tools/cldr/cldr-to-icu/src/main/java/org/unicode/icu/tool/cldrtoicu/Cldr2IcuCliOptions.java b/icu/tools/cldr/cldr-to-icu/src/main/java/org/unicode/icu/tool/cldrtoicu/Cldr2IcuCliOptions.java
+index d9b46014438..9ff85b38d86 100644
+--- a/icu/tools/cldr/cldr-to-icu/src/main/java/org/unicode/icu/tool/cldrtoicu/Cldr2IcuCliOptions.java
++++ b/icu/tools/cldr/cldr-to-icu/src/main/java/org/unicode/icu/tool/cldrtoicu/Cldr2IcuCliOptions.java
+@@ -380,7 +380,9 @@ private void validateEnvironment() {
+
+         if (!new File(icuDir).isDirectory()
+                 || ! new File(icuDir, "icu4c").isDirectory()
+-                || ! new File(icuDir, "icu4j").isDirectory()
++                // MSFT-Change: microsoft/icu fork is icu4c-only; no icu4j source tree.
++                // The Maven dependency on icu4j (used by TransformsMapper) is still
++                // resolved from ~/.m2 at runtime, so source-tree absence is harmless.
+                 || ! new File(icuDir, "tools/cldr/cldr-to-icu").isDirectory()
+                 || ! new File(icuDir, "tools/cldr/cldr-to-icu/pom.xml").isFile()) {
+             System.err.println("The `" + icuDir + "` directory does not look like a valid icu root.");

icu/icu4c/source/common/common.vcxproj

@@ -41,7 +41,8 @@
       <PrecompiledHeaderOutputFile>$(OutDir)/icuuc.pch</PrecompiledHeaderOutputFile>
       <AssemblerListingLocation>$(OutDir)/</AssemblerListingLocation>
       <ObjectFileName>$(OutDir)/</ObjectFileName>
-      <ProgramDataBaseFileName>$(OutDir)/icuuc.pdb</ProgramDataBaseFileName>
+      <!-- MSFT-Change: Include the ICU major version number in the PDB filename. -->
+      <ProgramDataBaseFileName>$(OutDir)/icuuc$(IcuMajorVersion).pdb</ProgramDataBaseFileName>
     </ClCompile>
     <Link>
       <!-- The icudt.lib is for U_ICUDATA_ENTRY_POINT -->
@@ -53,24 +54,34 @@
   <ItemDefinitionGroup Condition="'$(Configuration)'=='Debug'">
     <ClCompile>
       <PreprocessorDefinitions>RBBI_DEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
-      <RuntimeLibrary>MultiThreadedDebugDLL</RuntimeLibrary>
+      <!-- MSFT-Change: Static link the VCRUNTIME, VCSTARTUP and STL, but dynamic link to the UCRT. -->
+      <RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
     </ClCompile>
     <Link>
       <OutputFile>..\..\$(IcuBinOutputDir)\icuuc$(IcuMajorVersion)d.dll</OutputFile>
-      <ProgramDatabaseFile>.\..\..\$(IcuLibOutputDir)\icuucd.pdb</ProgramDatabaseFile>
+      <!-- MSFT-Change: Include the ICU major version number in the PDB filename. -->
+      <ProgramDatabaseFile>.\..\..\$(IcuLibOutputDir)\icuuc$(IcuMajorVersion)d.pdb</ProgramDatabaseFile>
       <ImportLibrary>..\..\$(IcuLibOutputDir)\icuucd.lib</ImportLibrary>
+      <!-- MSFT-Change: This forces dynamic linking of the UCRT. -->
+      <IgnoreSpecificDefaultLibraries>libucrtd.lib;libucrt.lib</IgnoreSpecificDefaultLibraries>
+      <AdditionalOptions>/DEFAULTLIB:ucrtd.lib %(AdditionalOptions)</AdditionalOptions>
     </Link>
   </ItemDefinitionGroup>
   <!-- Options that are common to all 'Release' project configurations -->
   <ItemDefinitionGroup Condition="'$(Configuration)'=='Release'">
     <ClCompile>
-      <RuntimeLibrary>MultiThreadedDLL</RuntimeLibrary>
+      <!-- MSFT-Change: Static link the VCRUNTIME, VCSTARTUP and STL, but dynamic link to the UCRT. -->
+      <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
       <FunctionLevelLinking>true</FunctionLevelLinking>
     </ClCompile>
     <Link>
       <OutputFile>..\..\$(IcuBinOutputDir)\icuuc$(IcuMajorVersion).dll</OutputFile>
-      <ProgramDatabaseFile>.\..\..\$(IcuLibOutputDir)\icuuc.pdb</ProgramDatabaseFile>
+      <!-- MSFT-Change: Include the ICU major version number in the PDB filename. -->
+      <ProgramDatabaseFile>.\..\..\$(IcuLibOutputDir)\icuuc$(IcuMajorVersion).pdb</ProgramDatabaseFile>
       <ImportLibrary>..\..\$(IcuLibOutputDir)\icuuc.lib</ImportLibrary>
+      <!-- MSFT-Change: This forces dynamic linking of the UCRT. -->
+      <IgnoreSpecificDefaultLibraries>libucrtd.lib;libucrt.lib</IgnoreSpecificDefaultLibraries>
+      <AdditionalOptions>/DEFAULTLIB:ucrt.lib %(AdditionalOptions)</AdditionalOptions>
     </Link>
   </ItemDefinitionGroup>
   <ItemGroup>
@@ -275,6 +286,7 @@
     <ClCompile Include="static_unicode_sets.cpp" />
     <ClCompile Include="restrace.cpp" />
     <ClCompile Include="fixedstring.cpp" />
+    <ClCompile Include="uprefs.cpp" />
     <ClInclude Include="localsvc.h" />
     <ClInclude Include="msvcres.h" />
     <ClInclude Include="pluralmap.h" />
@@ -394,6 +406,7 @@
     <ClInclude Include="capi_helper.h" />
     <ClInclude Include="restrace.h" />
     <ClInclude Include="fixedstring.h" />
+    <ClInclude Include="uprefs.h" />
   </ItemGroup>
   <ItemGroup>
     <ResourceCompile Include="common.rc" />

icu/icu4c/source/common/common_uwp.vcxproj

@@ -409,6 +409,7 @@
     <ClCompile Include="static_unicode_sets.cpp" />
     <ClCompile Include="restrace.cpp" />
     <ClCompile Include="fixedstring.cpp" />
+    <ClCompile Include="uprefs.cpp" />
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="localsvc.h" />
@@ -529,6 +530,7 @@
     <ClInclude Include="capi_helper.h" />
     <ClInclude Include="restrace.h" />
     <ClInclude Include="fixedstring.h" />
+    <ClInclude Include="uprefs.h" />
   </ItemGroup>
   <ItemGroup>
     <ResourceCompile Include="common.rc" />

icu/icu4c/source/common/putil.cpp

@@ -66,6 +66,7 @@
 #include "locmap.h"
 #include "ucln_cmn.h"
 #include "charstr.h"
+#include "uprefs.h"
 
 /* Include standard headers. */
 #include <stdio.h>
@@ -1794,10 +1795,37 @@ The leftmost codepage (.xxx) wins.
     return posixID;
 
 #elif U_PLATFORM_USES_ONLY_WIN32_API
-#define POSIX_LOCALE_CAPACITY 64
     UErrorCode status = U_ZERO_ERROR;
     char *correctedPOSIXLocale = nullptr;
 
+#if UCONFIG_USE_WINDOWS_PREFERENCES_LIBRARY 
+
+    int32_t neededBufferSize = uprefs_getBCP47Tag(nullptr, 0, &status);
+    MaybeStackArray<char,ULOC_FULLNAME_CAPACITY> windowsLocale(neededBufferSize, status);
+    int32_t length = uprefs_getBCP47Tag(windowsLocale.getAlias(), neededBufferSize, &status);
+
+    if (length > 0) // If length is 0, then the call to uprefs_getBCP47Tag failed.
+    {
+        // Now normalize the resulting name
+        correctedPOSIXLocale = static_cast<char *>(uprv_malloc(length * 2));
+        /* TODO: Should we just exit on memory allocation failure? */
+        if (correctedPOSIXLocale)
+        {
+            int32_t posixLen = uloc_canonicalize(windowsLocale.getAlias(), correctedPOSIXLocale, length * 2, &status);
+            if (U_SUCCESS(status))
+            {
+                *(correctedPOSIXLocale + posixLen) = 0;
+                gCorrectedPOSIXLocale = correctedPOSIXLocale;
+                gCorrectedPOSIXLocaleHeapAllocated = true;
+                ucln_common_registerCleanup(UCLN_COMMON_PUTIL, putil_cleanup);
+            }
+            else
+            {
+                uprv_free(correctedPOSIXLocale);
+            }
+        }
+    }
+#else
     // If we have already figured this out just use the cached value
     if (gCorrectedPOSIXLocale != nullptr) {
         return gCorrectedPOSIXLocale;
@@ -1839,11 +1867,11 @@ The leftmost codepage (.xxx) wins.
         }
 
         // Now normalize the resulting name
-        correctedPOSIXLocale = static_cast<char *>(uprv_malloc(POSIX_LOCALE_CAPACITY + 1));
+        correctedPOSIXLocale = static_cast<char *>(uprv_malloc(length * 2));
         /* TODO: Should we just exit on memory allocation failure? */
         if (correctedPOSIXLocale)
         {
-            int32_t posixLen = uloc_canonicalize(modifiedWindowsLocale, correctedPOSIXLocale, POSIX_LOCALE_CAPACITY, &status);
+            int32_t posixLen = uloc_canonicalize(modifiedWindowsLocale, correctedPOSIXLocale, length * 2, &status);
             if (U_SUCCESS(status))
             {
                 *(correctedPOSIXLocale + posixLen) = 0;
@@ -1857,6 +1885,7 @@ The leftmost codepage (.xxx) wins.
             }
         }
     }
+#endif
 
     // If unable to find a locale we can agree upon, use en-US by default
     if (gCorrectedPOSIXLocale == nullptr) {

icu/icu4c/source/common/sources.txt

@@ -163,6 +163,7 @@ unistr_titlecase_brkiter.cpp
 unorm.cpp
 unormcmp.cpp
 uobject.cpp
+uprefs.cpp
 uprops.cpp
 ures_cnv.cpp
 uresbund.cpp

icu/icu4c/source/common/ucln_cmn.cpp

@@ -36,8 +36,9 @@ static cleanupFunc *gLibCleanupFunctions[UCLN_COMMON];
  The cleanup order is important in this function.
  Please be sure that you have read ucln.h
  ************************************************/
+// MSFT-Change: Make u_cleanup a no-op for the Windows OS ICU version.
 U_CAPI void U_EXPORT2
-u_cleanup()
+uprv_u_cleanup()
 {
     UTRACE_ENTRY_OC(UTRACE_U_CLEANUP);
     icu::umtx_lock(nullptr);     /* Force a memory barrier, so that we are sure to see   */
@@ -52,6 +53,31 @@ u_cleanup()
 /*#endif*/
 }
 
+U_CAPI void U_EXPORT2
+u_cleanup()
+{
+// When ICU is built as an OS component for Windows, we make the public function u_cleanup
+// effectively a no-op because of the following:
+//  - It is not thread-safe and *forcefully* unloads the ICU data file.
+//  - The ICU library can simultaneously be used by other threads when this happens,
+//    either by Windows.Globalization, or by sorting functions when ICU sorting is default.
+//  - This means an App can call the function at any time, which will cause random crashes. :(
+//
+// We don't completely remove the functionality though, as we still want/need to be able to
+// unload resources when the ICU DLL is unloaded. Instead we make it a private (uprv) function
+// so that the combined DLL can still call it, and we don't export it in the DEF file.
+//
+// Note: We don't unconditionally do this though, as we don't want to alter the behavior of the
+//   public function when ICU when used/consumed in a Nuget package (for example).
+//
+#if defined(ICU_DATA_DIR_WINDOWS)
+    ((void)0); // no-op.
+    return;
+#else
+    uprv_u_cleanup();
+#endif
+}
+
 U_CAPI void U_EXPORT2 ucln_cleanupOne(ECleanupLibraryType libType) 
 {
     if (gLibCleanupFunctions[libType])

icu/icu4c/source/common/ucmndata.h

@@ -30,8 +30,12 @@
 #include "unicode/udata.h"
 #include "umapfile.h"
 
-
-#define COMMON_DATA_NAME U_ICUDATA_NAME
+// MSFT-Change: In the Windows OS ICU build, we only have one data package, and we use a versionless name in filename.
+#if defined(ICU_DATA_DIR_WINDOWS)
+#  define COMMON_DATA_NAME "icudtl"
+#else
+#  define COMMON_DATA_NAME U_ICUDATA_NAME
+#endif
 
 typedef struct  {
     uint16_t    headerSize;

icu/icu4c/source/common/udata.cpp

@@ -809,6 +809,16 @@ openCommonData(const char *path,          /*  Path from OpenChoice?          */
  *----------------------------------------------------------------------*/
 static UBool extendICUData(UErrorCode *pErr)
 {
+// MSFT-Change: For the Windows OS build of ICU, we only have one data file
+// and we don't use the extended data at all. We make this function a no-op
+// in order to save a few cycles for perf, but more importantly so that
+// we don't try to load a versioned data file (ex: icudt78l.dat) after
+// already loading the non-versioned common data file.
+#if defined(ICU_DATA_DIR_WINDOWS)
+    (void)pErr; // suppress unused variable.
+    return false;
+#endif
+
     UDataMemory   *pData;
     UDataMemory   copyPData;
     UBool         didUpdate = false;

icu/icu4c/source/common/unicode/putil.h

@@ -42,6 +42,9 @@
  * functions may have to be re-implemented.
  */
 
+//IGNORE_WINDOWS_HEADERS_START
+// MSFT-Change: The Windows OS version of ICU uses a single fixed data file.
+
 /**
  * Return the ICU data directory. 
  * The data directory is where common format ICU data files (.dat files)
@@ -112,7 +115,7 @@ U_CAPI const char * U_EXPORT2 u_getTimeZoneFilesDirectory(UErrorCode *status);
 U_CAPI void U_EXPORT2 u_setTimeZoneFilesDirectory(const char *path, UErrorCode *status);
 #endif  /* U_HIDE_INTERNAL_API */
 
-
+// MSFT-TODO: Should these be considered for Windows?
 /**
  * @{
  * Filesystem file and path separator characters.
@@ -134,6 +137,7 @@ U_CAPI void U_EXPORT2 u_setTimeZoneFilesDirectory(const char *path, UErrorCode *
 #   define U_FILE_ALT_SEP_STRING "/"
 #   define U_PATH_SEP_STRING ":"
 #endif
+//IGNORE_WINDOWS_HEADERS_END
 
 /** @} */
 

icu/icu4c/source/common/unicode/uchar.h

@@ -49,6 +49,10 @@ typedef struct USet USet;
 
 U_CDECL_BEGIN
 
+//IGNORE_WINDOWS_HEADERS_START
+// MSFT-Change: The value of these macros can change at runtime, so the API u_getUnicodeVersion
+// should be used instead of any version macro.
+
 /*==========================================================================*/
 /* Unicode version number                                                   */
 /*==========================================================================*/
@@ -63,6 +67,8 @@ U_CDECL_BEGIN
  */
 #define U_UNICODE_VERSION "17.0"
 
+//IGNORE_WINDOWS_HEADERS_END
+
 /**
  * \file
  * \brief C API: Unicode Properties

icu/icu4c/source/common/unicode/uconfig.h

@@ -45,6 +45,10 @@
  * @stable ICU 2.4
  */
 
+//IGNORE_WINDOWS_HEADERS_START
+// MSFT-Change: Since these are compile time settings, it doesn't make sense to 
+// load a user config header in the Windows OS SDK version.
+
 /**
  * If this switch is defined, ICU will attempt to load a header file named "uconfig_local.h"
  * prior to determining default settings for uconfig variables.
@@ -55,6 +59,8 @@
 #include "uconfig_local.h"
 #endif
 
+//IGNORE_WINDOWS_HEADERS_END
+
 /**
  * \def U_DEBUG
  * Determines whether to include debugging code.
@@ -379,6 +385,9 @@
 #   define UCONFIG_MSGPAT_DEFAULT_APOSTROPHE_MODE UMSGPAT_APOS_DOUBLE_OPTIONAL
 #endif
 
+//IGNORE_WINDOWS_HEADERS_START
+// MSFT-Change: We always use the OS LCID mapping API for the Windows OS build of ICU.
+
 /**
  * \def UCONFIG_USE_WINDOWS_LCID_MAPPING_API
  * On platforms where U_PLATFORM_HAS_WIN32_API is true, this switch determines
@@ -391,6 +400,24 @@
 #   define UCONFIG_USE_WINDOWS_LCID_MAPPING_API 1
 #endif
 
+/**
+ * \def UCONFIG_USE_WINDOWS_PREFERENCES_LIBRARY 
+ *   On Windows platforms (ie: U_PLATFORM_HAS_WIN32_API is true), this switch enables ICU to
+ *   detect additional user preferences by setting BCP47 Unicode extension within the default locale.
+ *   This includes information such as calendar, currency, hour cycle, among others.
+ *
+ *   If this switch is off (or set to 0) then the default behavior of only detecting the language
+ *   and country/region occurs. 
+ *
+ *   For example, the default locale may be detected as "es-MX-u-hc-h24", instead of "es-MX",
+ *   if the user has selected a 24 hour clock option.
+*/
+#ifndef UCONFIG_USE_WINDOWS_PREFERENCES_LIBRARY 
+#   define UCONFIG_USE_WINDOWS_PREFERENCES_LIBRARY  1
+#endif
+
+//IGNORE_WINDOWS_HEADERS_END
+
 /* i18n library switches ---------------------------------------------------- */
 
 /**

icu/icu4c/source/common/unicode/unistr.h

@@ -60,6 +60,9 @@ class Edits;
 
 U_NAMESPACE_END
 
+//IGNORE_WINDOWS_HEADERS_START
+// MSFT-Change: Hiding the @internal API below, since we don't expose the C++ UnicodeString.
+
 // Not #ifndef U_HIDE_INTERNAL_API because UnicodeString needs the UStringCaseMapper.
 /**
  * Internal string case mapping function type.
@@ -77,6 +80,8 @@ UStringCaseMapper(int32_t caseLocale, uint32_t options,
                   icu::Edits *edits,
                   UErrorCode &errorCode);
 
+//IGNORE_WINDOWS_HEADERS_END
+
 U_NAMESPACE_BEGIN
 
 class Locale;               // unicode/locid.h

icu/icu4c/source/common/unicode/utypes.h

@@ -104,6 +104,11 @@
 
 /** @} */
 
+//IGNORE_WINDOWS_HEADERS_START
+// MSFT-Change: For the Windows OS version of ICU, it doesn't make sense to expose these
+// constants which are only for loading the main ICU data file. We also don't 
+// support using a data DLL either, so omit them from the Windows SDK header.
+
 /*===========================================================================*/
 /* ICUDATA naming scheme                                                     */
 /*===========================================================================*/
@@ -191,6 +196,8 @@
 #endif
 #endif  /* U_HIDE_INTERNAL_API */
 
+//IGNORE_WINDOWS_HEADERS_END
+
 /**
  * \def NULL
  * Define NULL if necessary, to nullptr for C++ and to ((void *)0) for C.