<<< Date Index >>>     <<< Thread Index >>>

Re: introduction / first question and special characters



 On Friday, September 9, 2005 at 1:38:44 PM +0200, Daniel Hertrich wrote:

> That one was sent with the GMX webmail interface. The last one with
> mutt. This one comes with webmail again.

    Hope of small progress vanished?! ;-(


> How do I determine if libc-builtin iconv should be used or external
> libiconv? I could try both and see which one works better. If I only
> knew how.

    I don't know. Perhaps someone else? "ldd /usr/local/bin/mutt" can
say you what shared libraries are linked.


>> an LC_CTYPE file inside locales directories?
> only in the directories:
> /hdd2/QtPalmtop/lib/locale/ja_JP.eucjp/
> /hdd2/QtPalmtop/lib/locale/ja_JP.utf8

    Good news! You just have to find a terminal with EUC-JP charset, and
learn Japanese: Problem solved. You might be surprised that umlauts are
double-width characters in these locales, though... ;-)


> What do these files do?

    They are libc private binary structures about the character set.
Printability, width, control codes, and so on. They are necessary for
the correct functionning of a bunch of libc functions (out of US-Ascii
which is the builtin default).


> Does it make sense to simply transfer a de_DE directory from my Linux
> desktop to the machine we are talking about (BTW it is a Zaurus
> SL-C3000 palmtop)?

    I send you privately a trio of 2.2.5 compiled de_DE locales to see.
The ARM processor is bigendian, right? BTW an strace of any simple
locale using command ("strace date") can show where you system expects
to open locales. Example mine shows with LANG=fr_FR.UTF-8@euro:

| open("/usr/lib/locale/fr_FR.UTF-8@euro/LC_IDENTIFICATION", O_RDONLY) = 3

    The important info is "/usr/lib/locale/" or elsewhere.


>> :set &charset ?charset
> It always shows us-ascii. What should it show instead, and why?

    It should show the charset used by the current locale. Example it
should show "euc-jp" for the "ja_JP.eucjp" locale. Because the libc
provides a function for this, used by Mutt each time you set &charset:
char *nl_langinfo(CODESET).


> Why the heck does mutt behave so strangely? All other programs have no
> problems... But they are not self-compiled by me.

    That may be a reason for the iconv clash, not for lacking locales.
Other apps may use hardcoded workarounds similar to the +LOCALES_HACK
that breaks for you. Example the "less" pager: It uses the locale by
default, and hardcoded charsets when LESSCHARSET or LESSCHARDEF are set.

    Note in general for an app, using the locale is The Good Way.
Workarounds are considered dirty hacks to be avoided if possible.


> Thanks a lot for your guidance. I appreciate it very much.

    You are very welcome. I hope we can solve this misery, and that all
this can be later usefull to all Zaurus users.


Bye!    Alain.
-- 
Hotmail users break umlauts for everyone else on a mailing list!
They should stop doing so immediately!
        « MSN considered HARMFUL » PCC CB on MU. © June 2002