A terminal can use Unicode characters doesn't mean it has to be able to decode UTF-* encoded data, it merely means that you can encode your Unicode characters and the terminal an recognize that encoding. For example, GB18030 is a Unicode 5.0 implementation, but itself is nether UTF-8 or UTF-16. And many legacy encodings support the charset required by tqdm, but the currently _supports_unicode doesn't take this in account.

A visual difference is that pip displays progress bar nicely on my terminal (LC_CTYPE=GBK) but tqdm only displays #. You can find out how pip detects Unicode terminal support in pip/utils/ui.py:_select_progress_class.

And here is how I adopt pip's approach for fixing tqdm in my code:

def _has_unicode(fp):
    try:
        u'\u2588\u2589'.encode(fp.encoding)
    except (UnicodeEncodeError, AttributeError):
        return False
    else:
        return True