Re: D68K: The next step...
Hi,
Hamish wrote:
[..]
> The upshot is that installing libc5_5.2.18-9.deb killed my machine; no
> new process could load the libc shared library. I eventually repaired
> myself by downloading a libc-5.3.12 image from uni-erlangen and
> installing it "manually". Most of the binaries on my system were
> linked against 5.0.9, other than the ones which I had to recompile for
> the new utmp format, which were linked against my 5.2.18.
>
> I don't know if others have been able to get that deb archive to work,
> but I thought that I'd point out my problem.
I don't know if the problems I went through during the last days are
related to yours, but I could imagine the right 'mix' of ld.so and libc
is/was your problem. I still did not really test the 'base system', but
I'm coming closer to it - the scripts to create the base.tar.gz already
work now, so the configuration steps done during installation (dinstall)
are the only step still missing.
What took me a while to find out was that when you want to run a command
in a chroot()ed environment, the 'new' / environment does not only have
to supply all required shared libraries, ld-linux.so and libdl, but also
the exact same versions (files, that is). Otherwise I always get a
segmentation fault. Ok, after a few thoughts: ld-linux is always 'running',
and if some pages of code are being brought in from the new root environment,
but the binary over there looks different from the one in the old /lib,
things can only go wrong.. ever seen 'text file is busy'? :}
Frank (hereby having said that he's still alive)
PS: I sent a mail to Peter Anvin (maintainer of the registered major/minor
device numbers) asking him to resolve our current minor id clash for
amigamouse vs. 'jbm'. Unfortunately he's on vacation until July 29th.
Reply to: