Bug#232445: Anyone listening?

[Hadmut Danisch <hadmut@danisch.de>]

On Fri, Feb 20, 2004 at 05:45:06PM +0100, Bernd S. Brentrup wrote:
> On Fri, Feb 20, 2004 at 04:37:43PM +0100, Hadmut Danisch wrote:
> > Just for curiosity: Is anyone reading this bug report?
> 
> Just for curiosity (and I an no dpkg maintainer): what makes
> you think dpkg behaves as you want it to if you're modifying
> it's database under its feet?


Bullshit! 

I did not modify anything at the database. This bug
occured on a system freshly installed from a plain 
debian mirror. I was installing several packages as usual
with apt-get install, which worked well. when installing
the meta-package "gnome", dpkg went into an endless loop.
Stopping it and calling "dpkg --configure --pending" brought
dpkg into that loop again and again.

Since this machine is to be used for an application in a closed 
room and without any network access, I made a copy of the 
dpkg directory for later analysis at home. I did not modify anything
in dpkg's database. 

When I was home, I started to analyze the problem with my other
machines and found, that the problem can be simulated by just copying 
dpkg into an image freshly generated with debootstrap.

This is just for debugging purposes, not the live system. It allows to 
debug the problem without having access to the machine.

And to answer your (silly) question: I used strace to verify that 
there is no inconsistency between the database and other data. dpkg
only reads the dpkg directory before going into the endless loop, 
and this directory is exactly the same as on the machine where the 
bug occured. So I did not "modify the database under it's feet", 
because the database has not been modified. It's an exact copy.
Same input - same bug.




> Do you care to close this bug yourself please?

I won't close this bug before this issue is resolved.

This is a critical bug, since it makes the whole system 
unusable. Ignorance doesn't help here.

Hadmut