Re: new module 'duplocale'

[Ludovic Courtès]

Hi,

Eric Blake <address@hidden> writes:

> According to Ludovic Courtès on 11/23/2009 4:01 PM:
>>>       const char *base_name = nl_langinfo (_NL_LOCALE_NAME (LC_CTYPE));
>>
>> This is not thread-safe but I guess there’s no other choice.
>
> As is the case with a number of gnulib modules.  But I have been trying to
> document that in the various files; for example, note how
> doc/posix-functions/openat.texi mentions that the replacement is not
> thread-safe.

Yes.  In this case it’s probably safe to say that it’s the programmer’s
responsibility to do proper locking around uses of the global locale.

>> On a related note, nl_langinfo_l(3) is broken in current glibc:
>>
>>   http://sourceware.org/bugzilla/show_bug.cgi?id=11009
>
> For once, I'm tending to agree with Uli - the use of the global locale is
> important for duplocale to create a one-off variant locale, but when it
> comes to all the *_l interfaces, if you want the global locale, then use
> what got returned by duplocale, rather than teaching all the _l interfaces
> to special case the placeholder constant for the global locale.

Hmm, yes.  I think it’d have been much easier like this:

  #define LC_GLOBAL_LOCALE (&_nl_global_locale)

But I guess the reason it’s not done this way is so qthat
‘_nl_global_locale’ can have ‘hidden’ linkage so that the non-_l
variants don’t pay the penalty.  Oh, well.

Thanks,
Ludo’.