Re: A locale question.

[Ulrich Drepper <drepper@cygnus.com>]

hjl@lucon.org (H.J. Lu) writes:

> 1. VSX4L0 only defines LC_CTYPE. Should LC_ALL work? I mean when a new
>    locale only defines the parts of LC_ALL, should it uses the rest
>    from the default one? It seems that the POSIX testsuite expects this
>    behavior. Who is correct?

A partly constructed locale can only be used if only this part of
requested.  I.e., one has to use setlocale(LC_CTYPE, ...) instead of
setlocale(LC_ALL, ...).

The localedef program is written in a way that it automatically
creates all locale files even if they have only the default valies in
them.

> 2. It makes up some value for int_curr_symbol. glibc doesn't like it.
>    And glibc complains some other things.

I don't understand what you say.

> localedef -c -f lang0.ascii -i lang0.src VSX4L0
> Computing table size for character classes might take a while... done
> localedef -c -f lang1.ascii -i lang1.src VSX4L1
> no definition of `UNDEFINED'

See libc/229.  It will be fixed with the rewrite of the locale stuff
in glibc 2.2 (I'm almost done with it).

> value of field `int_curr_symbol' in category `LC_MONETARY' does not correspond to a valid name in ISO 4217

If the value does not correspond to a legal value the test suite is
buggy.  POSIX specifies that all names must come from this standard.
So, tell me what value is used and I'll check whether it is missing
from the list of legal values.

> field `grouping' in category `LC_NUMERIC' not defined
> field `yesexpr' in category `LC_MESSAGES' undefined
> field `noexpr' in category `LC_MESSAGES' undefined

I have to look it up whether this fields are mandatory or not.

-- 
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Cygnus Solutions `--' drepper at cygnus.com   `------------------------