Closed jbryan closed 8 years ago
:D This is terrible. Thank you for the bug report!
It looks like with probability 1 in 2 million (making some assumptions)---i.e., if the fractional seconds part is > 999999.5us---then some dodgy rounding happens and it incorrectly produces "...HH:MM:SS.1". I rewrote the bit that splits a timestamp into seconds and microseconds.
I believe I've fixed it in 78afdfc346293cfe934c3fc80c347e69691ffd05, and I've released version 0.7 with that change.
On certain floating point timestamps, I observe the following (tested in python 3.4 on 64bit ubuntu):
Other values of t that produce the exception are: