Uniscribe ScriptShape character not found in Simplified Arabic Windows Svr 2012 or W10

Question

I have an app which must convert strings out of rtf files to be rendered in pdf. I have a problem with Simplified Arabic Bold, it works in Windows 7, but fails in Windows 10 or W.Srv 2012.

A very simple word in Arabic: غ، دهو

I select the "simplified Arabic" font as returned by GDI+ ToNative() method. ScriptShape return the correct glyph Ids in Win7, but in Win10 the glyphIds returned are not in the CMAP table. For instance GlyphId 678 cannot be found, so i cannot add the character in the rendering of the PDF.

A more detailed explanation:

Here is the RTF reduced to the minimal expression to reproduce the problem:

{\rtf1\fbidis\ansi\ansicpg1252\deff0\nouicompat\deflang1033
  {\fonttbl{\f0\froman\fprq2\fcharset178 Simplified Arabic;}{\f1\fnil Source Sans Pro;}} 
   \pard\ltrpar\cf1\b\f0\rtlch\fs32\lang1025\'db\'a1 \'cf\'e5\'e6 \cf0\b0\f1\ltrch\fs18\lang1033\par
}

In order to generate the atoms for PDF, I first get the System.Drawing.Font("Simplified Arabic", 16, bold).

Then I call GetFontData(hDC, 0,0, buffer, length) and get the table TagGlyf data, constructing a dictionary <ushort, chart> from the GlyphIdArray[] of the tableCMap.Win (i can provide my code if you need it)

Then i call

Uniscribe.ScriptItemize(text, text.Length, items.Length, ref scriptControl, ref scriptState, items, ref itemsCount);

After getting the items, i call Uniscribe.ScriptLayout() to order the runs according to BiDi levels.

Then I select the Font in the current hDc and for each run call ScriptShape() which return the GlyphIds of the runs.

Using the Win7 Simplified Arabic (even in Win10) i get the following glyphs: glyphs

{ushort[9]}
    [0]: 3
    [1]: 194
    [2]: 395
    [3]: 174
    [4]: 3
    [5]: 157
    [6]: 185

and all are correctly matching a char that can i send to the PDF. However, when i replace and use the SimplifiedArabic Bold from W10 or WinSvr 2012, I get

glyphs {

ushort[9]}
    [0]: 3
    **[1]: 678**
    [2]: 395
    [3]: 174
    [4]: 3
    [5]: 157
    [6]: 185
    [7]: 0
    [8]: 0

You can see that the glyphs[1] in Win10 font is 678, but I cant find that Id in my dictionary from the CMap Table.

If I manually set the glyphs[1] to 194, it works ok rendering the correct character. So I don't know why ScriptShape gets confused about which GlyphId use with the new font.

(Of course if i install manually the old Win7 font in the Win10 box, it works perfectly)

Any suggestions?

thanks. df