While fixing the issue, I noticed that there's some slightly inconsistencies with how TzDateTimes that falls in the middle of an overlap are handled. For negative offsets, the earlier time is chosen, for positive offsets, the later time is chosen. I'm not sure whether this should be fixed since it might break backwards compatibility.
This PR fixes #164.
While fixing the issue, I noticed that there's some slightly inconsistencies with how
TzDateTime
s that falls in the middle of an overlap are handled. For negative offsets, the earlier time is chosen, for positive offsets, the later time is chosen. I'm not sure whether this should be fixed since it might break backwards compatibility.