I noticed that when my message is 160 characters but it contains the char € (or any symbols from extended table), the encoded_length is 161 as it should be, since the € char counts for two. However, in the make_parts function from gsm.py, the condition for knowing if the message should be split or not is based on len(text) which returns 160. To me, the condition should be based on len(encoded_text)

Edit : I noticed it because of submit_sm_resp failed and the ESME told me I wasn't splitting the message although it was 161 characters.