[Libreoffice-commits] core.git: Branch 'libreoffice-4-2' - vcl/win

[Herbert Dürr]

 vcl/win/source/gdi/winlayout.cxx |   17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

New commits:
commit b66d2073ff11b9306bc77ed91dabf9bcf7d29c4d
Author: Herbert Dürr <hdu at apache.org>
Date:   Thu Mar 27 16:07:37 2014 +0000

    Related: #i124516# handle bad surrogate pairs gracefully on Windows
    
    When running into invalid Unicode surrogate pairs the text layout code on
    Windows ran into massive problems like crashes. This change detects the
    situation of an invalid surrogate pair and falls back to treat it as
    a simple character instead of requesting a complex glyph fallback.
    
    (cherry picked from commit 913f1fc4b1362f6e91595af5ae10c4cba79fd355)
    
    Change-Id: I2988f4b64061d0a5df211f6f0f04b1f235fcd6a5
    (cherry picked from commit 67688d3118b1a361d5dbdaa78e918815c163d75c)

diff --git a/vcl/win/source/gdi/winlayout.cxx b/vcl/win/source/gdi/winlayout.cxx
index f240f17..b3fa345 100644
--- a/vcl/win/source/gdi/winlayout.cxx
+++ b/vcl/win/source/gdi/winlayout.cxx
@@ -386,12 +386,19 @@ bool SimpleWinLayout::LayoutText( ImplLayoutArgs& rArgs )
         bool bSurrogate = ((nCharCode >= 0xD800) && (nCharCode <= 0xDFFF));
         if( bSurrogate )
         {
-            if( nCharCode >= 0xDC00 ) // this part of a surrogate pair was already processed
+            // ignore high surrogates, they were already processed with their low surrogates
+            if( nCharCode >= 0xDC00 )
                 continue;
-            nCharCode = 0x10000 + ((pCodes[0] - 0xD800) << 10) + (pCodes[1] - 0xDC00);
-    }
+            // check the second half of the surrogate pair
+            bSurrogate &= (0xDC00 <= pCodes[1]) && (pCodes[1] <= 0xDFFF);
+            // calculate the UTF-32 code of valid surrogate pairs
+            if( bSurrogate )
+                nCharCode = 0x10000 + ((pCodes[0] - 0xD800) << 10) + (pCodes[1] - 0xDC00);
+            else // or fall back to a replacement character
+                nCharCode = '?';
+        }
 
-        // get the advance width for the current UCS-4 code point
+        // get the advance width for the current UTF-32 code point
         int nGlyphWidth = mrWinFontEntry.GetCachedGlyphWidth( nCharCode );
         if( nGlyphWidth == -1 )
         {
@@ -409,7 +416,7 @@ bool SimpleWinLayout::LayoutText( ImplLayoutArgs& rArgs )
         mpGlyphAdvances[ i ] = nGlyphWidth;
         mnWidth += nGlyphWidth;
 
-        // remaining codes of surrogate pair get a zero width
+        // the second half of surrogate pair gets a zero width
         if( bSurrogate && ((i+1) < mnGlyphCount) )
             mpGlyphAdvances[ i+1 ] = 0;