On Mon, Dec 18, 2006 at 12:16:24PM -0500, Bill Nottingham wrote:
Michal Jaegermann (michal(a)harddata.com) said:
> Cute indeed. :-)
> AFAICS my network modules are still in place. I just
> restored my ifcfg*.bak files to original names, did
> '/etc/init.d/network restart', and I was back in business.
> It would be more interesting exercise if I had only a network
> access to a stricken box.
The symptoms that you describe are roughly similar to what would
happen if you removed your network cards, booted once, and then
rebooted with them in.
I could not have done that without a screwdriver unless those
network interfaces were misdetected by something. It did not
happen before or after though - AFAICT.
So, the question would be why did that
code trigger in your case.
I am still in a dark. Digging through all of that I found something
interesting though. The following construct:
( LC_ALL=C; [[ ! "0" =~ '^[0-9A-Za-z_]*$' ]] && echo error ||
echo ok )
prints 'error' with bash-3.2-1.fc7 from rawhide but 'ok' with
bash-3.1-16.1 (FC6) and bash-3.1-9.fc5.1 (FC5).
In case anybody wonders a test of that sort is used at least in
/etc/sysconfig/network-scripts/ifup-aliases resulting in "invalid
alias number" error messages if you are trying to bring up an
interface while using 'ifup ifcf-eth...' with a corresponding
configuration file for a specific alias interface does not have any
I can only marvel what other tests (possibly silently) fail, or
failed, courtesy of bash. OTOH I got that bash version on my
system only yesterday and right now I could not tell what was
really a behaviour when I was hit by the original problem.
I guess that I better put something about bash into bugzilla.