The nti() function that converts a number field from a tar header to a number
failed to decode GNU tar specific base-256 fields. I also added support for
decoding and encoding negative base-256 number fields.

The nti() function that converts a number field from a tar header to a number
failed to decode GNU tar specific base-256 fields. I also added support for
decoding and encoding negative base-256 number fields.

encoding, instead of the locale encoding.

instead of the locale encoding.

because wcsftime() doesn't format time zone correctly.

wcsftime() doesn't format time zone correctly.

Also fix the bz2 module, which suffered from the same problem.

Also fix the bz2 module, whose classes used the same algorithm.

Use PyUnicode_IS_ASCII instead of PyUnicode_IS_COMPACT_ASCII, so the following
test can be removed:

   PyUnicode_DATA(op) == (((PyCompactUnicodeObject *)(op))->utf8)

Add also a special case if the result is an empty string.

Thanks to the PEP 3151, select.error (which is just an alias to OSError) has
now an error attribute.

"except" works properly.