[PATCH] Money_get::do_get: first create, then destroy

[Benjamin Kosnik]

This is fine thanks. Actually, it's a nice patch, testsuite entry, etc. I 
love that kind of patch.

Do you know if this kind of thing happens in other, 
related facets (num_get, time_get) or the sister money facet (money_put?)
I usually try to do a full audit for regressions like this. If
you can, you might want to glance around the other facets to see if this
same class of regression could exist elsewhere. 

> P.S. I'm afraid we are still not doing always the right thing when the sign is
> longer than 1 char. I will tackle this next.

Great.

The en_HK locale tests exercise a lot of (pathalogical/odd) behavior WRT
sign issues, so make sure that you are testing with more than the "C" 
locale... you've probably already noticed the testsuite entries, but I 
thought I'd give you a heads up anyway.

thanks again,
benjamin


> 2002-02-04  Paolo Carlini  <pcarlini@unitus.it>
> 
>         * include/bits/locale_facets.tcc (money_get::do_get(string)):
>         First build a tentative returned string, then, only if the
>         parsing succeeds, copy it into the string passed by reference.
>         * testsuite/22_locale/money_get_members_char.cc: Add test06.
>         * testsuite/22_locale/money_get_members_wchar_t.cc: Add test06.