Just found another flub, when an n/mdash entity was at beginning or end of line, the entity was being changed, but the count wasn't getting added to (as the extras/offset check wasn't getting a match), so the tags went out of alignment with the text.
Hopefully this should be it. (famous last words...)
Edit:
If you don't want it to do any of the (m/n)dash or ellipsis entities, you can comment out the following lines (add a # to beginning of the line) in the smarten.py file
32, 33, 34, 35, 36 and 56, 57, 58
Edit 2:
Yep, spoke too soon.
Just updated it, missed removing a couple of characters in the ndash replace line.
If you've already downloaded this new version, can you change line 57 in smarten.py to:
Code:
text = re.sub(r'(?<=[^-])--(?=[^-])', r'–', text, flags=re.M) # ndash