resolved rhbz#919020 Basic CDbl() and CSng() scan localized number

Code wrongly replaced decimal separator with '.' and used atof() to scan
the number string which itself may be localized on *iX systems but not
on Windows. Hence on *iX the numbers may had been truncated where on
Windows they were not.

Additionally made this work with Unicode separators if defined for the
locale, removed the awkward conversion to ASCII byte string and
eliminated use of toupper() and isdigit() calls. Also eliminated a
possible buffer overflow.

Change-Id: Ic3f791601c321ef1386fe51a5244d3dacac3d908

basic/source/sbx/sbxscan.cxx

@@ -55,108 +55,131 @@ void ImpGetIntntlSep( sal_Unicode& rcDecimalSep, sal_Unicode& rcThousandSep )
     rcThousandSep = rData.getNumThousandSep()[0];
 }
 
+inline bool ImpIsDigit( sal_Unicode c )
+{
+    return '0' <= c && c <= '9';
+}
+
+/** NOTE: slightly differs from strchr() in that it does not consider the
+    terminating NULL character to be part of the string and returns bool
+    instead of pointer, if character is 0 returns false.
+ */
+bool ImpStrChr( const sal_Unicode* p, sal_Unicode c )
+{
+    if (!c)
+        return false;
+    while (*p)
+    {
+        if (*p++ == c)
+            return true;
+    }
+    return false;
+}
+
+bool ImpIsAlNum( sal_Unicode c )
+{
+    return (c < 128) ? isalnum( static_cast<char>(c) ) : false;
+}
+
 // scanning a string according to BASIC-conventions
-// but exponent may also be a D, so data type is SbxDOUBLED
+// but exponent may also be a D, so data type is SbxDOUBLE
 // conversion error if data type is fixed and it doesn't fit
 
 SbxError ImpScan( const OUString& rWSrc, double& nVal, SbxDataType& rType,
                   sal_uInt16* pLen, bool bAllowIntntl, bool bOnlyIntntl )
 {
-    OString aBStr( OUStringToOString( rWSrc, RTL_TEXTENCODING_ASCII_US ) );
-
-    char cIntntlComma, cIntntl1000;
-    char cNonIntntlComma = '.';
-
-    sal_Unicode cDecimalSep, cThousandSep = 0;
+    sal_Unicode cIntntlDecSep, cIntntlGrpSep;
+    sal_Unicode cNonIntntlDecSep = '.';
     if( bAllowIntntl || bOnlyIntntl )
     {
-        ImpGetIntntlSep( cDecimalSep, cThousandSep );
-        cIntntlComma = (char)cDecimalSep;
-        cIntntl1000 = (char)cThousandSep;
+        ImpGetIntntlSep( cIntntlDecSep, cIntntlGrpSep );
+        if( bOnlyIntntl )
+            cNonIntntlDecSep = cIntntlDecSep;
     }
-
     else
     {
-        cIntntlComma = cNonIntntlComma;
-        cIntntl1000 = cNonIntntlComma;
+        cIntntlDecSep = cNonIntntlDecSep;
+        cIntntlGrpSep = 0;  // no group separator accepted in non-i18n
     }
 
-    if( bOnlyIntntl )
-    {
-        cNonIntntlComma = cIntntlComma;
-        cIntntl1000 = (char)cThousandSep;
-    }
-
-    const char* pStart = aBStr.getStr();
-    const char* p = pStart;
-    char buf[ 80 ], *q = buf;
+    const sal_Unicode* const pStart = rWSrc.getStr();
+    const sal_Unicode* p = pStart;
+    OUStringBuffer aBuf( rWSrc.getLength());
     bool bRes = true;
     bool bMinus = false;
     nVal = 0;
     SbxDataType eScanType = SbxSINGLE;
-    while( *p &&( *p == ' ' || *p == '\t' ) ) p++;
+    while( *p == ' ' || *p == '\t' )
+        p++;
     if( *p == '-' )
-        p++, bMinus = true;
-    if( isdigit( *p ) ||( (*p == cNonIntntlComma || *p == cIntntlComma ||
-            *p == cIntntl1000) && isdigit( *(p+1 ) ) ) )
+    {
+        p++;
+        bMinus = true;
+    }
+    if( ImpIsDigit( *p ) || ((*p == cNonIntntlDecSep || *p == cIntntlDecSep ||
+                    (cIntntlDecSep && *p == cIntntlGrpSep)) && ImpIsDigit( *(p+1) )))
     {
         short exp = 0;
-        short comma = 0;
+        short decsep = 0;
         short ndig = 0;
         short ncdig = 0;    // number of digits after decimal point
-        OStringBuffer aSearchStr("0123456789DEde");
-        aSearchStr.append(cNonIntntlComma);
-        if( cIntntlComma != cNonIntntlComma )
-            aSearchStr.append(cIntntlComma);
+        OUStringBuffer aSearchStr("0123456789DEde");
+        aSearchStr.append(cNonIntntlDecSep);
+        if( cIntntlDecSep != cNonIntntlDecSep )
+            aSearchStr.append(cIntntlDecSep);
         if( bOnlyIntntl )
-            aSearchStr.append(cIntntl1000);
-        const char* pSearchStr = aSearchStr.getStr();
-        while( strchr( pSearchStr, *p ) && *p )
+            aSearchStr.append(cIntntlGrpSep);
+        const sal_Unicode* const pSearchStr = aSearchStr.getStr();
+        const sal_Unicode pDdEe[] = { 'D', 'd', 'E', 'e', 0 };
+        while( ImpStrChr( pSearchStr, *p ) )
         {
-            if( bOnlyIntntl && *p == cIntntl1000 )
+            aBuf.append( *p );
+            if( bOnlyIntntl && *p == cIntntlGrpSep )
             {
                 p++;
                 continue;
             }
-
-            if( *p == cNonIntntlComma || *p == cIntntlComma )
+            if( *p == cNonIntntlDecSep || *p == cIntntlDecSep )
             {
-                // always insert '.' so that atof works
                 p++;
-                if( ++comma > 1 )
+                if( ++decsep > 1 )
                     continue;
-                else
-                    *q++ = '.';
+                // Use the separator that is passed to stringToDouble()
+                aBuf[ p - pStart ] = cIntntlDecSep;
             }
-            else if( strchr( "DdEe", *p ) )
+            else if( ImpStrChr( pDdEe, *p ) )
             {
                 if( ++exp > 1 )
                 {
-                    p++; continue;
+                    p++;
+                    continue;
                 }
-                if( toupper( *p ) == 'D' )
+                if( *p == 'D' || *p == 'd' )
                     eScanType = SbxDOUBLE;
-                *q++ = 'E'; p++;
-
-                if( *p == '+' )
-                    p++;
-                else
-                if( *p == '-' )
-                    *q++ = *p++;
+                aBuf[ p - pStart ] = 'E';
+                p++;
             }
             else
             {
-                *q++ = *p++;
-                if( comma && !exp ) ncdig++;
+                p++;
+                if( decsep && !exp )
+                    ncdig++;
             }
-            if( !exp ) ndig++;
+            if( !exp )
+                ndig++;
         }
-        *q = 0;
 
-        if( comma > 1 || exp > 1 )
+        if( decsep > 1 || exp > 1 )
             bRes = false;
 
-        if( !comma && !exp )
+        OUString aBufStr( aBuf.makeStringAndClear());
+        rtl_math_ConversionStatus eStatus = rtl_math_ConversionStatus_Ok;
+        sal_Int32 nParseEnd = 0;
+        nVal = rtl::math::stringToDouble( aBufStr, cIntntlDecSep, cIntntlGrpSep, &eStatus, &nParseEnd );
+        if( eStatus != rtl_math_ConversionStatus_Ok || nParseEnd != aBufStr.getLength() )
+            bRes = false;
+
+        if( !decsep && !exp )
         {
             if( nVal >= SbxMININT && nVal <= SbxMAXINT )
                 eScanType = SbxINTEGER;
@@ -164,49 +187,65 @@ SbxError ImpScan( const OUString& rWSrc, double& nVal, SbxDataType& rType,
                 eScanType = SbxLONG;
         }
 
-        nVal = atof( buf );
-        ndig = ndig - comma;
+        ndig = ndig - decsep;
         // too many numbers for SINGLE?
         if( ndig > 15 || ncdig > 6 )
             eScanType = SbxDOUBLE;
 
         // type detection?
-        if( strchr( "%!&#", *p ) && *p ) p++;
+        const sal_Unicode pTypes[] = { '%', '!', '&', '#', 0 };
+        if( ImpStrChr( pTypes, *p ) )
+            p++;
     }
     // hex/octal number? read in and convert:
     else if( *p == '&' )
     {
         p++;
         eScanType = SbxLONG;
-        const char *cmp = "0123456789ABCDEF";
+        OUString aCmp( "0123456789ABCDEFabcdef" );
         char base = 16;
         char ndig = 8;
-        char xch  = *p++;
-        switch( toupper( xch ) )
+        switch( *p++ )
         {
-            case 'O': cmp = "01234567"; base = 8; ndig = 11; break;
-            case 'H': break;
-            default : bRes = false;
+            case 'O':
+            case 'o':
+                aCmp = "01234567";
+                base = 8;
+                ndig = 11;
+                break;
+            case 'H':
+            case 'h':
+                break;
+            default :
+                bRes = false;
         }
-        long l = 0;
-        int i;
-        while( isalnum( *p ) )
+        const sal_Unicode* const pCmp = aCmp.getStr();
+        while( ImpIsAlNum( *p ) )    /* XXX: really munge all alnum also when error? */
         {
-            char ch = sal::static_int_cast< char >( toupper( *p ) );
+            sal_Unicode ch = *p;
+            if( ImpStrChr( pCmp, ch ) )
+            {
+                if (ch > 0x60)
+                    ch -= 0x20;     // convert ASCII lower to upper case
+                aBuf.append( ch );
+            }
+            else
+                bRes = false;
             p++;
-            if( strchr( cmp, ch ) ) *q++ = ch;
-            else bRes = false;
         }
-        *q = 0;
-        for( q = buf; *q; q++ )
+        OUString aBufStr( aBuf.makeStringAndClear());
+        long l = 0;
+        for( const sal_Unicode* q = aBufStr.getStr(); bRes && *q; q++ )
         {
-            i =( *q & 0xFF ) - '0';
-            if( i > 9 ) i -= 7;
-            l =( l * base ) + i;
+            int i = *q - '0';
+            if( i > 9 )
+                i -= 7;     // 'A'-'0' = 17 => 10, ...
+            l = ( l * base ) + i;
             if( !ndig-- )
                 bRes = false;
         }
-        if( *p == '&' ) p++;
+        if( *p == '&' )
+            p++;
         nVal = (double) l;
         if( l >= SbxMININT && l <= SbxMAXINT )
             eScanType = SbxINTEGER;