Re: [RFC] First draft of the POSIX locale::name patch

[Benjamin Kosnik <bkoz at redhat dot com>]

> 1- _M_c_cats(__cat) contains the names of the various
>    C categories present in recent glibc: arguably a
>    rough approach.
> 
>    1a- I have no idea how to automatically adapt it (at
>        configure time?!?) to different libc... Is this needed
>        or in fact ;) the named locales machinery presently
>        works well only for glibc anyway.
> 
>    1b- When the 6 additional categories stabilized? This
>        is relevant for the new testcases too (some should be
>        conditionalized to glibc >= 2.1 or >= 2.0??)

Solution enclosed. Perhaps this should be two separate arrays, not one
combined one.

> 2- Is there a clean way to avoid strtok() in
>    locale::_Impl::_Impl(const char*, size_t) not changing
>    to much the infrastructure?

I'll look at this next

> 3- Should we consider the possibility of names longer than
>    20 chars (see the _M_names array)?

fixed already in the patch you'd sent me

> 4- Is it ok to use the non ANSI strdup() in
>    locale::locale(const char *) and in
>    locale::_Impl::_Impl(const char*, size_t)? Otherwise I
>    could easily replace it with malloc + strlen calls.

I think so

Attachment: p.20021004
Description: Binary data