At Fri, 11 Oct 2002 11:34:57 -0400, Ben Collins wrote: > On Fri, Oct 11, 2002 at 10:52:24PM +0900, GOTO Masanori wrote: > > At Fri, 11 Oct 2002 06:00:40 -0700, > > Jeff Bailey wrote: > > > I will hopefully be hacking some of our build code, though. > > > Cross-compile cases seem to be broken right now, which limits my > > > ability to test hurd-i386. Also that might give me a chance to fix > > > the error report I saw about installing libc6-686. > > > > libc6-586/686 is still broken, so I'm fixing it (but I'm fighting my > > HDD RAID because it's crashed). The problem is 'ld-linux.so.2' does > > not put on 586/686 directory. In addition, adding like > > > > ln -sf /lib/686/ld-linux.so.2 /lib > > > > is needed for postinst/prerm. After fix, I think i386 architecture is > > ready to release 2.3.1-1 deb package :) > > Oh, fuck no. Do not do that. There's nothing much to be gained. What > could possibly be the problem with this? The whole reason I added the > /etc/ld.so.nohwcap is so we didn't need to mess with the dynamic linker > symlink. It's asking for trouble, believe me.
I understand why you think it's bad. ld-linux.so.2 is as 'the only one', we don't include it for optimization package. BTW, i586/i686 does not work, many apps gets segv with -opt package (I'm investigating). I think releasing glibc 2.3.1 without -opt is the most important matter, and releasing -opt (for i586/i686/alpha/sparcv9/hurd-i*/ mips/...) is the next item. Dropping -opt for the first glibc 2.3.1 package is the good direction for me. Any comments are welcome! Regards, -- gotom
