BUG: invalid construction from repr of dt-aware index

[jreback]

thought we had an issue about this....

The data is converted from string and is tz-aware, then the dtype convert is also applied.

In [1]: dt = pd.date_range('20160101', periods=1, tz='US/Eastern')

In [2]: dt
Out[2]: DatetimeIndex(['2016-01-01 00:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq='D')

In [3]: r = DatetimeIndex(['2016-01-01 00:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq='D')

In [4]: r
Out[4]: DatetimeIndex(['2016-01-01 05:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq='D')

In [5]: dt.equals(r)
Out[5]: False

[mroeschke]

After #21822 this now raises

In [3]: r = DatetimeIndex(['2016-01-01 00:00:00-05:00'], dtype='datetime64[ns, US/Eastern]', freq='D')
TypeError: data is already tz-aware pytz.FixedOffset(-300), unable to set specified tz: US/Eastern

Which I argue is correct since the user is attempting to localize the datetime with two methods

• dtype
• utcoffset