First of all I must say, that yesterday I was tired and wanted to sleep, so I was inadequate.
Today I understood, that everything is done according to standards, but Marshall Banana was wrong about charset. If reply is in HEX or UCS2, then it is printed hex encoded without any conversion. If it was coded in default GSM alphabet n900 automatically decodes it to current terminal equipment charset.
By default CP437 is used, not CP850. I don't think it is a good choice, because it doesn't contain greek characters, so not all characters present in GSM encoding can be coded in CP437. But we can switch terminal equipment to GSm encoding with
at+cscs="GSM"
command. Then our code for decoding would work correctly. I'll make special version us ussdquery.py which does this trick and ask you to test it.
I was too nerwous yesterday.
There are many ways to debug. Of course you can try blindly switch options in widget or making low level queries. But the fastest an easiest way is to run widget from console. In most cases this will be enough to find out, where definitely error
occurred. I tried to say this many times.