codecvt<wchar_t, char, mbstate_t> (was RE: Libstdc++ Copyright Assignment)

[Benjamin Kosnik]

On Fri, 10 Jan 2003 15:10:30 -0000
Pétur Runólfsson <peturr02@ru.is> wrote:

>Looks more like most of the work. The problem is that everything must
>be tested with many different locales and encodings. Each locale will
>require it's own test data and expected results will differ between
>locales and systems. Available locales differ between systems.

More on this later....

Have you looked at testsuite/22_locale, and 

codecvt.cc                       codecvt_members_unicode_wchar_t.cc
codecvt_members_char_char.cc     codecvt_members_wchar_t_char.cc
codecvt_members_unicode_char.cc

?

The locale bits need to be broken down into

22_locale/codecvt_members/char/01.cc
...
22_locale/codecvt_members/char/0x.cc

and

22_locale/codecvt_members/wchar_t/01.cc
...
22_locale/codecvt_members/wchar_t/0x.cc

at which point some of the XPASS bits can go away with more intelligent
construction of the test list.

 -benjamin