From 2c17722ec8a15d93ef9ec4a4ff390a0db9a34d9d Mon Sep 17 00:00:00 2001
From: Vadim Zeitlin <vadim@wxwidgets.org>
Date: Tue, 1 May 2007 14:11:06 +0000
Subject: [PATCH] fix wxStringOutputStream::Write() in Unicode build when the
 output overlaps a boundary between UTF-8 characters (closes bug 1701426)

git-svn-id: https://svn.wxwidgets.org/svn/wx/wxWidgets/trunk@45733 c3d73ce0-8a6f-49c7-b76d-6d57e0e08775
---
 docs/changes.txt       |  2 ++
 include/wx/sstream.h   |  8 +++++++
 src/common/sstream.cpp | 49 +++++++++++++++++++++++++++++++++++++++---
 3 files changed, 56 insertions(+), 3 deletions(-)

diff --git a/docs/changes.txt b/docs/changes.txt
index d254faba40..80f192640e 100644
--- a/docs/changes.txt
+++ b/docs/changes.txt
@@ -151,6 +151,8 @@ wxX11:
 All:
 
 - Fix bug in wxFileConfig when recreating a group (Steven Van Ingelgem)
+- Fix wxStringOutputStream::Write() in Unicode build when the argument
+  overlaps UTF-8 characters boundary
 - Account for lines without newline at the end in wxExecute()
 
 All (Unix):
diff --git a/include/wx/sstream.h b/include/wx/sstream.h
index 02f878681b..f8e75790ab 100644
--- a/include/wx/sstream.h
+++ b/include/wx/sstream.h
@@ -61,6 +61,9 @@ public:
     // The stream will write data either to the provided string or to an
     // internal string which can be retrieved using GetString()
     wxStringOutputStream(wxString *pString = NULL)
+#if wxUSE_UNICODE_WCHAR
+        : m_unconv(0)
+#endif // wxUSE_UNICODE_WCHAR
     {
         m_str = pString ? pString : &m_strInternal;
         m_pos = m_str->length() / sizeof(wxChar);
@@ -90,6 +93,11 @@ private:
     wxMBConv m_conv;
 #endif
 
+#if wxUSE_UNICODE_WCHAR
+    // unconverted data from the last call to OnSysWrite()
+    wxMemoryBuffer m_unconv;
+#endif // wxUSE_UNICODE_WCHAR
+
     DECLARE_NO_COPY_CLASS(wxStringOutputStream)
 };
 
diff --git a/src/common/sstream.cpp b/src/common/sstream.cpp
index a16ea1b24a..f92916a667 100644
--- a/src/common/sstream.cpp
+++ b/src/common/sstream.cpp
@@ -153,9 +153,52 @@ size_t wxStringOutputStream::OnSysWrite(const void *buffer, size_t size)
 {
     const char *p = wx_static_cast(const char *, buffer);
 
-    // append the input buffer (may not be null terminated - thus 
-    // the literal length
-    m_str->Append(wxString(p, m_conv, size));
+#if wxUSE_UNICODE_WCHAR
+    // the part of the string we have here may be incomplete, i.e. it can stop
+    // in the middle of an UTF-8 character and so converting it would fail; if
+    // this is the case, accumulate the part which we failed to convert until
+    // we get the rest (and also take into account the part which we might have
+    // left unconverted before)
+    const char *src;
+    size_t srcLen;
+    if ( m_unconv.GetDataLen() )
+    {
+        // append the new data to the data remaining since the last time
+        m_unconv.AppendData(p, size);
+        src = m_unconv;
+        srcLen = m_unconv.GetDataLen();
+    }
+    else // no unconverted data left, avoid extra copy
+    {
+        src = p;
+        srcLen = size;
+    }
+
+    wxWCharBuffer wbuf(m_conv.cMB2WC(src, srcLen, NULL /* out len */));
+    if ( wbuf )
+    {
+        // conversion succeeded, clear the unconverted buffer
+        m_unconv = wxMemoryBuffer(0);
+
+        *m_str += wbuf;
+    }
+    else // conversion failed
+    {
+        // remember unconverted data if there had been none before (otherwise
+        // we've already got it in the buffer)
+        if ( src == p )
+            m_unconv.AppendData(src, srcLen);
+
+        // pretend that we wrote the data anyhow, otherwise the caller would
+        // believe there was an error and this might not be the case, but do
+        // not update m_pos as m_str hasn't changed
+        return size;
+    }
+#else // !wxUSE_UNICODE_WCHAR
+    // no recoding necessary, the data is supposed to already be in UTF-8 (if
+    // supported) or ASCII otherwise
+    m_str->append(p, size);
+#endif // wxUSE_UNICODE_WCHAR/!wxUSE_UNICODE_WCHAR
 
     // update position
     m_pos += size;
-- 
2.45.2