Date: Sat, 22 Dec 2001 10:36:10 +0200 From: Alexey Zelkin <phantom@FreeBSD.org> To: "Andrey A. Chernov" <ache@nagual.pp.ru> Cc: cvs-committers@FreeBSD.org, cvs-all@FreeBSD.org Subject: Re: cvs commit: src/share/msgdef Makefile src/etc Makefile src/etc/mtree BSD.usr.dist src/lib/libc/locale lmessages.c Message-ID: <20011222103610.B16292@ark.cris.net> In-Reply-To: <20011221194928.GB1820@nagual.pp.ru>; from ache@nagual.pp.ru on Fri, Dec 21, 2001 at 10:49:29PM %2B0300 References: <200112211314.fBLDE2b42673@freefall.freebsd.org> <20011221151756.GC98626@nagual.pp.ru> <20011221184745.A91245@ark.cris.net> <20011221184818.GA1468@nagual.pp.ru> <20011221211354.A8081@ark.cris.net> <20011221194928.GB1820@nagual.pp.ru>
next in thread | previous in thread | raw e-mail | index | archive | help
hi, On Fri, Dec 21, 2001 at 10:49:29PM +0300, Andrey A. Chernov wrote: > > > hierarchy' when mtree tries to make directory over existen file. > > Please take a look on patches committed. It deals with 'make distrib-dirs' > > case. > > What about 'make hierarchy' case? Per src/Makefile.inc1 'make hierarchy' just calls 'make distrib-dirs'. > > To get short language names work we need to bring iconv first. Isn't it ? > > This issue is not related to iconv. To get short names work, we need > locale aliasing expansion work (to get long _name_ (not contents) by short > one). I.e. besides that setlocale() must return full name, we need > 'locale' shell utility, as described in POSIX, for shell scripts. > > But all this is unrelated to libintl. Libintl reduces long names to short > ones automatically. I mean programs that installs their .po files into > short directory names. But how it should handle showing ru_RU[.KOI8-R]'s .po file in case if user has locale set to ru_RU.CP866 for example ? > > I don't see a reason to do it in this way. > > Do you prefer to keep short and long names both? I don't see a reason to add new incompatibility by adding new directory hierarchy for libintl's po files. > > > Security issues may arise too (evil-made .po files). What about NLSPATH > > .po files should be generated as part of buildworld process. > > And then replaced at runtime by evil .po files using search patch > mechanism. If there is no such mechanism at all in libintl, all ports will > be forced to install their .po files into /usr/share/locale. As I can understand it's side effect of any functionality which is supporting search path mechanizms. What exactly you mean under term "evil" ? > > > analog in libintl? Is there a way to work f.e. with both /usr/share/locale > > > and /usr/local/share/locale? > > Let's back to this issue when libintl import time come. > > I think just otherwise: all this issues must be resolved _before_ any > libintl-specific changes sneaked in. Maybe import time even never come. This is not libintl specific change. This is change which should simplify future changes. As example was mentioned 'libintl'. Since FreeBSD never has support for LC_MESSAGES and changes still not MFC'ed it should not cause any breakages. I did it as last round of structural changes in preparation of MFC new locale stuff. All 'libintl' specific changes should be submitted to review and discussed before doing anything. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe cvs-all" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20011222103610.B16292>