Re: new AIX failure

[Paolo Carlini <pcarlini at unitus dot it>]

David Edelsohn wrote:

>	What confuses me is that the "C" locale test was added to many of
>the 22_locale files.  Only codecvt_members_char_char.cc fails.  If the
>other (non-wchar) files are not disabled, something more subtle is going
>on.
>
Indeed, in principle I can understand your perplexities...
However, let us consider the complete list of files hosting the new tests:

22_locale/codecvt_members_char_char.cc
22_locale/codecvt_members_wchar_t_char.cc
22_locale/collate_members_char.cc
22_locale/collate_members_wchar_t.cc
22_locale/ctype_members_char.cc
22_locale/ctype_members_wchar_t.cc
22_locale/messages_members_char.cc
22_locale/money_get_members_char.cc
22_locale/money_get_members_wchar_t.cc
22_locale/money_put_members_char.cc
22_locale/money_put_members_wchar_t.cc
22_locale/moneypunct_members_char.cc
22_locale/moneypunct_members_wchar_t.cc
22_locale/num_get_members_char.cc
22_locale/num_get_members_wchar_t.cc
22_locale/num_put_members_char.cc
22_locale/num_put_members_wchar_t.cc
22_locale/numpunct_members_char.cc
22_locale/numpunct_members_wchar_t.cc
22_locale/time_get_members_char.cc
22_locale/time_get_members_wchar_t.cc
22_locale/time_put_members_char.cc
22_locale/time_put_members_wchar_t.cc


Then, let us consider which ones of the above are XFAILed on 
non-gnu-locale model systems (just look for the dejagnu directives at 
the beginning of them:

22_locale/collate_members_char.cc
22_locale/collate_members_wchar_t.cc
22_locale/ctype_members_char.cc
22_locale/ctype_members_wchar_t.cc
22_locale/messages_members_char.cc
22_locale/money_get_members_char.cc
22_locale/money_get_members_wchar_t.cc
22_locale/money_put_members_char.cc
22_locale/money_put_members_wchar_t.cc
22_locale/moneypunct_byname.cc
22_locale/moneypunct_members_char.cc
22_locale/moneypunct_members_wchar_t.cc
22_locale/num_get_members_char.cc
22_locale/num_get_members_wchar_t.cc
22_locale/num_put_members_char.cc
22_locale/num_put_members_wchar_t.cc
22_locale/numpunct_byname.cc
22_locale/numpunct_members_char.cc
22_locale/numpunct_members_wchar_t.cc
22_locale/time_get_members_char.cc
22_locale/time_get_members_wchar_t.cc
22_locale/time_put_members_char.cc
22_locale/time_put_members_wchar_t.cc

What remains is only:

22_locale/codecvt_members_char_char.cc
22_locale/codecvt_members_wchar_t_char.cc

The second one in fact is /completely/ (new tests included) wrapped in 
an #ifdef _GLIBCPP_USE_WCHAR_T and, is suppose, is trivially passing on AIX.

Do you agree with my analysis? Benjamin, do you agree too? Does is make 
sense to both of you to fix all of the recent tests as you suggested before?

Thanks,
Paolo.