Re: correct value for LANG variable
On Sat, Jan 22, 2005 at 04:52:45PM +0100, Bj?rn K?nig wrote: I am a bit confused. Neither de_DE.ISO-8859-1 nor de_DE.ISO8859-1 work properly in all cases. setenv LANG de_DE.ISO8859-1 echo abcdef uvwxyz | tr '[a-z]' '[A-Z]' ABCDEF ?WXY?] For better or worse, this is not the correct way to perform case conversion in non-ASCII locales on FreeBSD 5 and later. See the COMPATIBILITY section of the tr(1) manpage for more information. setenv LANG de_DE.ISO-8859-1 echo abcdef uvwxyz | tr '[a-z]' '[A-Z]' ABCDEF UVWXYZ perl perl: warning: Setting locale failed. [lot of indignation about locale settings] FreeBSD uses the name de_DE.ISO8859-1 instead of de_DE.ISO-8859-1. Most base system utilities don't complain if LANG or the related LC_* variables are invalid. I am suspicious that de_DE.UTF-8 would work correct with most software. Is it the only solution to don't use localization currently at all or should it be my ambition to fix this? What should be the correct value? I would prefer the notation ISO-8859-1, but even the almighty perl does not. Setting LANG to de_DE.ISO8859-1 should work. The alternative is UTF-8, which is understood by most text-processing utilities in the base system (ls, wc, grep, sed, ...). However, at this stage neither sh nor vi work correctly with UTF-8, so you would need to replace these with (for example) bash and vim. Tim ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: correct value for LANG variable
On Saturday, 22. January 2005 16:52, Björn König wrote: I am a bit confused. Neither de_DE.ISO-8859-1 nor de_DE.ISO8859-1 work properly in all cases. setenv LANG de_DE.ISO8859-1 echo abcdef uvwxyz | tr '[a-z]' '[A-Z]' ABCDEF ÚWXYÝ] This is the correct locale name, but don't use tr like that, you can't rely on that to work correctly with charsets which aren't symmetric in loweruppercase characters (it used to work at some point for the ISO8859-1 [5] locales but ceased after a POSIX related modification to tr). Use tr [:lower:] [:upper:] instead. -- ,_, | Michael Nottebrock | [EMAIL PROTECTED] (/^ ^\) | FreeBSD - The Power to Serve | http://www.freebsd.org \u/ | K Desktop Environment on FreeBSD | http://freebsd.kde.org pgpILG8oAmJGV.pgp Description: PGP signature
Re: correct value for LANG variable
Tim Robbins wrote: For better or worse, this is not the correct way to perform case conversion in non-ASCII locales on FreeBSD 5 and later. See the COMPATIBILITY section of the tr(1) manpage for more information. Thanks a lot for this hint. I will pay more attention to this section. Regards Björn ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: correct value for LANG variable
Michael Nottebrock schrieb: Use tr [:lower:] [:upper:] instead. I was wondering about a third-party script which always worked for years with Solaris. I just didn't realized that this issue concerns a difference between SysV and BSD. I tried to set another LANG string and it worked, but now I see that it will work this way with an arbitrary string, because it's invalid. ;-) Björn ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
Re: correct value for LANG variable
On Sat, Jan 22, 2005 at 06:29:29PM +0100, Bj?rn K?nig wrote: Michael Nottebrock schrieb: Use tr [:lower:] [:upper:] instead. I was wondering about a third-party script which always worked for years with Solaris. I just didn't realized that this issue concerns a difference between SysV and BSD. The current behaviour of range expressions in tr was required by the Single Unix Specification Version 2 (SUSv2). Version 3 loosened this requirement by making these expressions unspecified in locales other than POSIX. This means you can't use tr with range expressions in any script intending to be portable without first setting LANG to POSIX. For reference, the systems I know of that sort range expressions according to the current locale are FreeBSD 5 (and later), AIX, and Tru64. The systems I know of that don't sort are FreeBSD 4 (and earlier), GNU (coreutils), and NetBSD. Tim ___ freebsd-stable@freebsd.org mailing list http://lists.freebsd.org/mailman/listinfo/freebsd-stable To unsubscribe, send any mail to [EMAIL PROTECTED]
