Can the new USE variable with name "diet" to be added? With it some of the
ebuilds (like qmail, djbdns etc.) can be compiled not with libc but with
dietlibc (http://www.fefe.de/dietlibc/) for size reducing.
not sure if we need USE, since we have virtual/glibc, but I dunno enough to say
I think dietlibc cannot replace glibc in whole. It is useful only for some
packages, not for the whole system. So USE var is better than virtual/glibc
I do not see the point. Sure, they will be smaller, but you will have an
extra package installed (dietlibc).
From the dietlibc page:
"The diet libc is a libc that is optimized for small size. It can be used to
create small statically linked binaries for Linux on alpha, arm, hppa, ia64,
i386, mips, s390, sparc, sparc64, ppc and x86_64"
Do you see the point - SMALL and STATICALLY LINKED. This means that finally
binaries, created with dietlibc does not require it. So after creating binaries
dietlibc can be removed, if it take too much space. You can put prodiced binaries
on floppy or something else and be sure, that they will not require some missing
I do understand the point out of a floppy distro, and maybe with a bootdisk
(although then qmail and djbdns do not really hold water for bootdisk), but
we are not a floppy distro.
And as I really cannot see the point in compiling stuff in the natural
run of things in Gentoo against dietlibc, except for floppy distro, or
for even less things, a bootdisk. For this reason I dont see adding another
USE flag for something that have no merit during the normal run of things,
meaning adding more work for us, meaning creating a bigger chance for new
and time consuming bugs.
If the one out of a thousand Gentoo users want to compile whatever against
dietlibc for his bootstiffy, he can easily just quickly edit the build, add
the required flag (not even 5 second job ? ), and then merge/package it.
Gentoo is supposed to be lean, mean, and speedy, but these days we just keep
on adding features/whatever because users are too lazy to just add it to
their profile/config or edit a build for two secs. We should include the
defaults that should be ok for 90% of the users, and the 2% *really* advanced
users who really want to do something at a wierd way, will not moan to edit
a script or two to change to suite themselfs.