with GS 8.53 the characters are missing wheras in Acrobat Professional 6.0 they are shown. Interestingly enough I saw no hints because of missing fonts and text marked in Acrobat and pasted in notepad with an adequate font was shown as all unknown characters. Single characters from the character table (windows utility) however were shown correctly. Possibly the characters are defined within the PDF and are mapped to "wrong" positions.
Created attachment 1915 [details] 10027.pdf Problem file.
I have determined that the characters _ARE_ being drawn into the character cache, so it does not seem to be a character mapping issue (those usually issue .notdef warnings).
The characters are being rendered into the cache, but the text position is incorrect when the cached char is to be 'blitted' onto the page so that the entire character is outside the clip box in effect when the text is being rendered. Soemthing is wrong w.r.t. the CTM and the Tm matrix resulting in the characters all being put near the lower left of the page, but there is a clipping box up where the characters are supposed to be. I can probably find this fairly quickly now that I've learned this much.
The Arabic characters are missing because they were not positioned correctly, in fact, they were outside the clipping box that was in effect as the text was painted. The problem was that there was a 'cm' (concatmatrix) PDF operation inside the BT/ET sections that painted the text. The correction involves updating the TextSaveMatrix when the 'cm' was inside a BT/ET block (which is when the TextSaveMatrix is 'known'. The patch can be viewed at: http://svn.ghostscript.com:8080/cgi-bin/viewcvs.cgi/ghostscript/trunk/gs/lib/pdf_draw.ps?rev=6458&r1=6446&r2=6458
Incomplete patch -- works for -r72, but doesn't for other scale factors. Test environment had other patches present which allowed this to work at all resolutions, but they were not committed as part of this fix. I will do so once those other patches are tested (parts of other fixes).
concatmatrix operand order was inverted causing incorrect transformation when resolution was not 72 dpi. This is now fixed in pdf_draw.ps rev. 6475