On Wed, 2007-04-18 at 12:00 +0200, Fabian Groffen wrote:
> On 04-04-2007 18:33:04 +0200, Michael Haubenwallner wrote:
<snip>
> > And this also works:
> > 
> > dev-libs/libiconv-1.11
> 
> keyworded now

Ah, thanks!

> 
> > Now you think wth I need libiconv on linux.
> > 
> > Well, I'm currently trying the profile on linux having ELIBC="linux"
> > instead of ELIBC="glibc", to avoid stray dependencies on glibc, because
> > I don't want to have any libc in prefix.
> 
> Sounds good.  Do you suggest using ELIBC=linux in prefix by default
> here?

Yes, at least as an option to be discussed.

<discussion on ELIBC="linux">

> 
> > IMO the easiest way is to say ELIBC="linux" - I've already
> > empty-remerged system using this profile.
> > 
> > Maybe virtual/libiconv could avoid a dependency on dev-libs/libiconv for
> > elibc_linux, but that's just an optimization.
> 
> Why avoid?  If you don't want to know about the installed glibc, you
> need libiconv if you do iconv stuff, right?

I treat this as elibc_linux++.

Some more questions to think of when deciding to use GNU libiconv on
linux (if answer='yes' then I'd say elibc_linux++):

Does uclibc lack (good) libiconv ?
Are there other libc's possible for linux than glibc and uclibc ?
Are there old glibc's around with bad libiconv implementations ?

If we do not use elibc_linux, and uclibc lacks libiconv, then maybe we
need to know of elibc_uclibc too in prefix...

</discussion on ELIBC="linux">

/haubi/


-- 
Salomon Automation GmbH - Friesachstraße 15 - A-8114 Friesach bei Graz
Sitz der Gesellschaft: Friesach bei Graz
UID-NR:ATU28654300 - Firmenbuchnummer: 49324 K
Firmenbuchgericht: Landesgericht für Zivilrechtssachen Graz

--
gentoo-alt@g.o mailing list