Darren Tucker wrote:
> The only thing to watch for there is for kernel-related things since
> you're obviously running on the same kernel in the chroot.

Certainly. And a good thing to note.

> This was the case with the descriptor passing bugs in Linux 2.0 kernels:
> the Debian folk used chroots as you described. configure found that
> descriptor passing worked fine (which it did, on the host's kernel)
> even though it didn't on the target. They eventually added some code
> to detect the buggy kernel versions at runtime.

Yes that can be a problem. But that is almost the same problem as
when you compile something for the local machine and then boot to a
different kernel on the same machine. Things that depend upon the
kernel might work differently. So what you say the Debian folks did,
which was to detect the situation at runtime, sounds like a best case
solution for things that depend upon the kernel.

I say almost the same problem because obviously most people don't lose
features when installing new kernels. But it does happen at times. I
often boot into older kernels in order to test something or to try to
recreate some particular configuration. Fortunately there are few
things that are sensitive to the kernel version.


openssh-unix-dev mailing list