to_hdf broken for timezone aware datetime64[ns] columns in pandas 0.17.0

[nickos556]

Saving DataFrame's to HDF format is broken for the new timezone aware datetime64[ns] datatype.
See the following example:

import pandas as pd
import numpy as np
from datetime import datetime

print "pandas version", pd.__version__

df1 = pd.DataFrame(data=np.random.randn(5,3), index=pd.DatetimeIndex(start=datetime.now(),freq='S',periods=5))
df1 = df1.reset_index() #Create a column of type datetime64[ns] (timezone naiive)
print df1.info()
df1.to_hdf('df.hdf', 'data_frame') #WORKS!!!

df2 = pd.DataFrame(data=np.random.randn(5,3), index=pd.DatetimeIndex(start=datetime.now(),freq='S',periods=5, tz='Australia/Sydney'))
df2 = df2.reset_index() #Create a column of type datetime64[ns, Australia/Sydney] (timezone AWARE)
print df2.info()
df2.to_hdf('df.hdf', 'data_frame') #BREAKS IN pandas 17.0!!!

[jreback]

fixed by #11628

and in 0.17.1