Re: Arla 0.42-RC2

[email protected] (Niels Möller) writes:
> I don't quite agree, I understand that I have to set up a full chroot
> environment if I want to exec new programs and stuff, but to have to
> do it also for chroot isn't very nice. If nothing else, it makes it a
> lot harder to use chroot in portable server programs portable, since
> the set names of the files you need to copy into the chroot can be
> very different between systems.
>
True. I doubt it is at all possible to figure out what is needed. libraries
can probably be figured out, but the complete set of files and helper
programs accessed is... "tricky".

> But I don't feel I'm the right person to argue about that with the
> glibc folks. Either someone else has to do that, or arla needs some
> workaround for the problem.
> 
Any ideas?

I doubt that one can argue from a standards standpoint, SUS2 says[1] about
chroot() "There is no portable use that an application could make of this
interface." 

If we have friends with commercial support from RedHat they may be able to
push it through that way.  I'd say forced preloading of "all libraries" is
reasonable given how common chroot() still is(?) in the real world.

The easy fix from our side would be to simply not use chroot().  Any
thoughts?

/t (no clue)

1. http://www.opengroup.org/onlinepubs/007908799/xsh/chroot.html