Re: Reverting to stable from test
David Z. Maze <dmaze@debian.org> wrote:
> Are you trying to build a kernel module? If so, are you explicitly
> giving it a path to the kernel headers, or are you letting it default
> to /usr/include/linux? The stuff in /usr/include/linux is almost
> certainly wrong and has a ~0% chance of corresponding to the kernel
> you're actually using; see /usr/share/doc/libc6/README.Debian.gz for
> details of what's going on, and why.
Actually, no. My applications are strictly user-land. There is only
one explicit reference to a "linux/" header (linux/pci.h) anywhere in
the code and that is an error (it should not be there).
The compile error I posted was just an example; there were may others
of a similarly extreme nature. Sadly I cannot recreate them as I have
in the course of the day tried removing a few things and reverting to
stable.
I have ended up with no development stuff at all. I baulked at removing
libc6 as a big long list of dependencies was presented, along with an
indication that it would be a bad idea. However, I get the feeling I
need to remove or replace libc6 as dselect refuses to do anything useful
as long as a conflict exists between libc6-dev (currently not installed)
depending upon libc6 2.1.3-18 (2.2.3-5 installed).
Is there an easy way out of this? I was just considering playing with
dpkg and its force-downgrade option, fully expecting that this would be
followed shortly after by a fresh installaion. Any advice?
I guess this will teach me not to touch anything marked "test" or
"unstable" if I don't know exactly what I'm doing.
Thanks for taking the trouble to reply.
Regards
--
William Morris
wrm@innovation-tk.com
Reply to: