tdf#133589 Numbertext: Use Win32API to avoid std::codecvt_utf8 bug

(cherry picked from commit 21a59b59d9a40ca32d91b05e62ffcd9aef8fd324
and commit ad792bccc0adfdd4be75f99170a92312924d371d)

Change-Id: I45c85db44c3dfd92e0929f66c8c95cb309c91e05
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/96894
Tested-by: Jenkins
Reviewed-by: Adolfo Jayme Barrientos <fitojb@ubuntu.com>

4 files changed, 86 insertions, 6 deletions

diff --git a/external/libnumbertext/MSVCNonBMPBug.patch1 b/external/libnumbertext/MSVCNonBMPBug.patch1
new file mode 100644
index 000000000000..8ced22165bcc
--- /dev/null
+++ b/external/libnumbertext/MSVCNonBMPBug.patch1
@@ -0,0 +1,69 @@
+MSVC's std::codecvt_utf8 has a bug converting non-BMP codepoints like U+10CFA.
+Use MultiByteToWideChar/WideCharToMultiByte instead on Windows.
+
+diff --git a/src/Numbertext.cxx b/src/Numbertext.cxx
+index 5f05b48579af..eb83e59f366f 100755
+--- a/src/Numbertext.cxx
++++ b/src/Numbertext.cxx
+@@ -7,6 +7,10 @@
+ #include <sstream>
+ #include <fstream>
+ 
++#ifdef _WIN32
++#include <Windows.h>
++#endif
++
+ #include "Numbertext.hxx"
+ 
+ #ifdef NUMBERTEXT_BOOST
+@@ -22,6 +26,14 @@
+ 
+ bool readfile(const std::string& filename, std::wstring& result)
+ {
++#ifdef _WIN32
++    std::ifstream ifs(filename);
++    if (ifs.fail())
++        return false;
++    std::stringstream ss;
++    ss << ifs.rdbuf();
++    result = Numbertext::string2wstring(ss.str());
++#else
+     std::wifstream wif(filename);
+     if (wif.fail())
+         return false;
+@@ -29,6 +44,7 @@ bool readfile(const std::string& filename, std::wstring& result)
+     std::wstringstream wss;
+     wss << wif.rdbuf();
+     result = wss.str();
++#endif
+     return true;
+ }
+ 
+@@ -99,7 +112,12 @@
+ 
+ std::wstring Numbertext::string2wstring(const std::string& s)
+ {
+-#ifndef NUMBERTEXT_BOOST
++#ifdef _WIN32
++    int nSize = MultiByteToWideChar(CP_UTF8, 0, s.c_str(), -1, nullptr, 0);
++    std::unique_ptr<wchar_t[]> wstr(new wchar_t[nSize]);
++    MultiByteToWideChar(CP_UTF8, 0, s.c_str(), -1, wstr.get(), nSize);
++    return wstr.get();
++#elif !defined NUMBERTEXT_BOOST
+     typedef std::codecvt_utf8<wchar_t> convert_type;
+     std::wstring_convert<convert_type, wchar_t> converter;
+     return converter.from_bytes( s );
+@@ -110,7 +128,12 @@
+ 
+ std::string Numbertext::wstring2string(const std::wstring& s)
+ {
+-#ifndef NUMBERTEXT_BOOST
++#ifdef _WIN32
++    int nSize = WideCharToMultiByte(CP_UTF8, 0, s.c_str(), -1, nullptr, 0, nullptr, nullptr);
++    std::unique_ptr<char[]> str(new char[nSize]);
++    WideCharToMultiByte(CP_UTF8, 0, s.c_str(), -1, str.get(), nSize, nullptr, nullptr);
++    return str.get();
++#elif !defined NUMBERTEXT_BOOST
+     typedef std::codecvt_utf8<wchar_t> convert_type;
+     std::wstring_convert<convert_type, wchar_t> converter;
+     return converter.to_bytes( s );
diff --git a/external/libnumbertext/UnpackedTarball_libnumbertext.mk b/external/libnumbertext/UnpackedTarball_libnumbertext.mk
index 1969dcf7d08d..d2efd2f852c3 100644
--- a/external/libnumbertext/UnpackedTarball_libnumbertext.mk
+++ b/external/libnumbertext/UnpackedTarball_libnumbertext.mk
@@ -15,4 +15,8 @@ $(eval $(call gb_UnpackedTarball_update_autoconf_configs,libnumbertext))
 
 $(eval $(call gb_UnpackedTarball_set_patchlevel,libnumbertext,1))
 
+$(eval $(call gb_UnpackedTarball_add_patches,libnumbertext, \
+    external/libnumbertext/MSVCNonBMPBug.patch1 \
+))
+
 # vim: set noet sw=4 ts=4:
diff --git a/lingucomponent/source/numbertext/numbertext.cxx b/lingucomponent/source/numbertext/numbertext.cxx
index bc13530d5250..556dc2ea7a90 100644
--- a/lingucomponent/source/numbertext/numbertext.cxx
+++ b/lingucomponent/source/numbertext/numbertext.cxx
@@ -22,6 +22,7 @@
 
 #include <osl/file.hxx>
 #include <tools/debug.hxx>
+#include <o3tl/char16_t2wchar_t.hxx>
 
 #include <sal/config.h>
 #include <cppuhelper/factory.hxx>
@@ -144,12 +145,20 @@ OUString SAL_CALL NumberText_Impl::getNumberText(const OUString& rText, const Lo
     if (!aCountry.isEmpty())
         aCode += "-" + aCountry;
     OString aLangCode(OUStringToOString(aCode, RTL_TEXTENCODING_ASCII_US));
+#if defined(_WIN32)
+    std::wstring sResult(o3tl::toW(rText.getStr()));
+#else
     OString aInput(OUStringToOString(rText, RTL_TEXTENCODING_UTF8));
-    std::wstring aResult = Numbertext::string2wstring(aInput.getStr());
-    bool result = m_aNumberText.numbertext(aResult, aLangCode.getStr());
+    std::wstring sResult = Numbertext::string2wstring(aInput.getStr());
+#endif
+    bool result = m_aNumberText.numbertext(sResult, aLangCode.getStr());
     DBG_ASSERT(result, "numbertext: false");
-    OString aResult2(Numbertext::wstring2string(aResult).c_str());
-    return OUString::fromUtf8(aResult2);
+#if defined(_WIN32)
+    OUString aResult(o3tl::toU(sResult.c_str()));
+#else
+    OUString aResult = OUString::fromUtf8(Numbertext::wstring2string(sResult).c_str());
+#endif
+    return aResult;
 #else
     return rText;
 #endif
diff --git a/sw/qa/extras/uiwriter/uiwriter.cxx b/sw/qa/extras/uiwriter/uiwriter.cxx
index 66137c9af662..2b8490b09135 100644
--- a/sw/qa/extras/uiwriter/uiwriter.cxx
+++ b/sw/qa/extras/uiwriter/uiwriter.cxx
@@ -7285,7 +7285,6 @@ void SwUiWriterTest::testTdf128860()
 #if ENABLE_LIBNUMBERTEXT
 void SwUiWriterTest::testTdf133589()
 {
-#if !defined(_WIN32)
     // Hungarian test document with right-to-left paragraph setting
     SwDoc* pDoc = createDoc("tdf133589.fodt");
     SwWrtShell* pWrtShell = pDoc->GetDocShell()->GetWrtShell();
@@ -7317,7 +7316,6 @@ void SwUiWriterTest::testTdf133589()
     nIndex = pWrtShell->GetCursor()->GetNode().GetIndex();
     sReplaced += u"𐳺𐳺𐳿𐳼𐳼 ";
     CPPUNIT_ASSERT_EQUAL(sReplaced, static_cast<SwTextNode*>(pDoc->GetNodes()[nIndex])->GetText());
-#endif
 }
 #endif