[RFC] moneypunct::frac_digits < 0 ?!?

[Paolo Carlini]

Hi,

speaking of corner cases, this is one: according to the standard,
frac_digits() returns an int... "of course" ;) in our current
implementation we don't deal gracefully with frac_digits() < 0 (*)

So, what shall we do? Consistency with other sections of the
standard (i.e., num_put::do_put for negative precision) and Note
244 seem to suggest that we should use by "default" the value 2...

Opinions?

Paolo.

(*) Before someone ;) says that the standard is plain wrong and we
must ignore the relevant paragraph: DR 374 NAD.